BI9999 PTY Report C1308785, A1317

1

To what extent is the distribution of Zostera noltii in

South and West Wales changing?

Laura Pratt

Biology

Leanne Cullen-Unsworth and Benjamin Jones

Project Seagrass, Sustainable Places Research Institute

Word Count: 6590

BI9999 PTY Report C1308785, A1317

2

Abstract

Seagrass meadows are declining globally at an unprecedented rate, despite their

valuable ecosystem services. Extensive loss has been observed in Europe

emphasising the importance of monitoring trends in the distribution of seagrass

species in order to monitor and conserve this fragile ecosystem. This study focuses

on the distribution of Zostera noltii in South and West Wales for which detailed

information on the spatial and temporal variations is sparse. An effective monitoring

method is therefore needed to quantify Z. noltii populations. The aim of this study

was to provide a fine-scale assessment into the historical and current known

distributions of Z. noltii and the magnitude of expansion/regression observed.

Developing accurate maps of seagrass extent is difficult, presenting challenges

including variable spatial distributions across a meadow, unfirm substrata and

changes in water clarity. Given these challenges this research utilised a range of

assessment methods and examined the potential use of a UAV (Unmanned Aerial

Vehicle) for assessing seagrass distribution. The extent, health and resilience of Z.

noltii were assessed throughout South and West Wales and data compared against

historical distributions. The outcomes showed the current spatial distribution of Z.

noltii, covering a known area of 73.3ha, with an overall estimated expansion of

2.1%yr-1±20.8. Sites in Pembrokeshire such as West Williamson displayed a

significantly higher seagrass condition with increased coverage, shoot density and

canopy height values observed. However, sites in Glamorgan mostly on Llanrhidian

Sands displayed a poorer Z. noltii health and signs of regression. Furthermore, in 5

of the locations, where Z. noltii had historically been recorded it was no longer

present most likely due to extensive anthropogenic impacts.

Keywords: Zostera noltii, UAV, remote sensing, Wales, Spatial heterogeneity

BI9999 PTY Report C1308785, A1317

3

Table of Contents

Page number

List of Figures………………………………………………………………………………..4

List of Boxes ………………………………………………………………………………...5

List of Tables ………………………………………………………………………………...6

1.0 Introduction……………………………………………………………………………...7

1.1 Z. noltii status.……………………………..……………………………………9

1.2 Z. noltii monitoring………………………..……………………………….……9

1.3 Unmanned aerial vehicles (UAVs) in Z. noltii monitoring…………………10

1.4 Aims and Objectives…………………………………………………………..11

2.0 Materials and Methods………………………………………………………………..13

2.1 Study sites……………………………………………………………………...13

2.2 Mapping of the intertidal Z. noltii meadow boundary………………………16

2.3 Drone Based data collection………………………………………………….16

2.4 Morphological measurements………………………………………………..20

2.5 Seed coring…………………………………………………………………….21

2.6 Data Analysis…………………………………………………………………..21

3.0 Results………………………………………………………………………………….24

3.1 Z. noltii extent…………………………………………………………………. 24

3.2 Drone analysis…………………………………………………………………30

3.3 Seagrass Morphometrics……………………………………………………..38

3.4 Historical change in Z. noltii morphometrics………………………………..41

3.5 Multivariate analysis of morphometric variables…………………………...42

3.6 Zostera Condition Index………………………………………………………43

3.7 Z. noltii morphometrics vs NDVI……………………………………………..45

3.8 Core Analysis…………………………………………………………………..45

3.9 Field observations……………………………………………………………..47

4.0 Discussion……………………………………………………………………………...49

4.1 Seagrass status and extent.………………………………………………….49

4.2 Affectability of drones as an alternative seagrass monitoring platform….54

4.3 Conclusion……………………………………………………………………..56

5.0 Acknowledgments………………………………………………………….………….57

6.0 References……………………………………………………………………………..58

Appendices (Supporting Information)……………………………………………………65

BI9999 PTY Report C1308785, A1317

4

List of Figures

Page number

Figure 1 Zostera noltii submerged in surface water located at Penclacwydd….……………...9

Figure 2 Geographical locations of Z. noltii (dwarf eelgrass) in Europe..…………………….10

Figure 3 Global decadal trends in seagrass extent, with sites either decreasing, increasing or displaying no change in the size of a meadow.……………………………………………….11

Figure 4 Map of the seagrass study sites visited along the South and West Wales coastline. Orange indicates sites that have a historical record of Z. noltii and green indicates sites that have been predicted to have suitable conditions for Zostera growth……….……….…..…….16

Figure 5 Example flight path of 3DR Solo at Garren Pill, superimposed on orthomosaic generated on Photoscan 1.2.0. Waypointed routes were plotted using Mission Planner and flown using pixhawk autopilot……………………………………………………………………...20

Figure 6 Example of overlapping images and computed camera positions recorded at A) Pwllcrochan and B) Garren Pill, C) fine- scale analysis at Angle Bay and D) lower altitude flight at Garren Pill, 2016…………………………………………………………………………...20

Figure 7 Steps used to create a three-dimensional model in Agisoft PhotoScan Professional………………………………………………………………………………………….21

Figure 8 The process of coring; 1) Cores taken randomly with a standard PVC seed corer (50mm diameter x 250mm long) and cap, pushed approximately 10cm deep in the sediment; 2) Sediment placed in stainless steel sieve (1-2mm mesh); 3) Sieved sediment using a little water and 4) Analysis of core content………………...…………………………………………..24

Figure 9 Differences in the area of seagrass recorded by 1) using a handheld GPS and 2) using transects, to assess different forms methodologies in monitoring larger Z. noltii meadows (A, Angle Bay and B, Penclacwydd)…………………………………………………..28

Figure 10 Non-linear regression model that can be used to predict the current Z. noltii extent based on historical (<2016) distributions…………………………………………………………29

Figure 11 Evolution of the spatial distribution of Z. noltii from 2013 to 2016, at sites showing signs of regression………………………………………………………………………………….30

Figure 12 Evolution of the spatial distribution of Z. noltii from 2013 to 2016, at sites showing signs of expansion………………………………………………………………………………….31

Figure 13 Differences in the quality of the aerial data recorded at Garren Pill, July 2016, on a customized 3DR Solo, flown at 15m using; A) RICOH GR II consumer grade compact camera, B) modified Xiaomi Yi Action Camera (RGB) and C) modified Xiaomi Yi Action Camera (NIR/Blue filtered)…………………………………………………………………………33

Figure 124 Differences in the area of Z. noltii at A) Pwllcrochan and B) Garron Pill, recorded using Garmin GPSmap 62 and forms of classification of the UAV data carried out on ArcMap 10.2.2..……………………………………………………………………………………………….34

Figure 15 Unsupervised ISO Cluster classification of true-colour mosaics at both A) Pwllcrochan (30m) and B) Garren Pill (50m)…………………………………………………….35

Figure 13 Supervised Maximum Likelihood image classification (MLC) of true-colour mosaics at both A) Pwllcrochan and B) Garren Pill……………………………………………..36

Figure 17 The process of manual classification used to assess Z. noltii boundaries at Pwllcrochan………………………………………………………………………………………….37

BI9999 PTY Report C1308785, A1317

5

Figure 18 Relative NDVI map generated from true-colour and NIR imagery at taken using a modified Xiaomi Yi Action Camera, at Garron Pill, July 2016……………………………….…38

Figure 19 Differences in the NDVI (Means±95%Cl) calculated for the different spectral classifications of the intertidal aerial data……………………………………………….………..39

Figure 20 Mean (±95%Cl) Z. noltii cover, shoot density, leaf length, leaf width, epiphyte cover and algae cover across all 10 study sites. Green indicates values that were recorded in a Z. noltii meadow that displayed signs of expansion, yellow no detectable change in the meadow size and red, a Z. noltii meadow with observed regression correlations in the area…………………………………………………………………………………………….……..42

Figure 21 Changes in the average Z. noltii cover (Means±95CI) and mean canopy height (mm) observed at Angle Bay from 1996 to 2000. Blue indicates historical values obtained from Hodges and Howe, 2007, yellow, Bunker, 2008, and green, Duggan-Edwards and Brazier, 2015, with the shoot density values obtained from the same data source………….43

Figure 22 Principal component analysis (PCA) of seagrass characteristics from 10 different study sites, spread along the South and West Wales coastline…………………………….....45

Figure 23 Non-linear regression model, estimating the relationship between the Normalised Difference Vegetation Index (NDVI) and dominant Z. noltii habitat characteristics. Red represents changes in Z. noltii cover, yellow shoot density and green canopy height………48

Figure 24 Mean (±95%Cl) Z. noltii seed density (m2) at both sites where seagrass was present (green) and sites where Zostera spp were absent (red)………………………………49

Figure 25 Area of intensive bait digging to the north of the intertidal Z. noltii bed in Angle Bay. Source: authors own, 2016…………………………………………………………………..50

Figure 26 Opportunistic macroalgae observed at West Williamson, from extensive algae cover of the Z. noltii bed. Source: authors own, 2016……………………………………….….50

Figure 27 Observed Z. noltii bleaching at Angle Bay, where the causality is unknown. Source: authors own, 2016………………………………………………………………………...50

Figure 28 Tractor marks observed at Zostera meadow in Porthdinllaen…………………….54

Figure 29 Aerial data recorded at Pwllcrochan, July 2016, where strong winds led to the model (DJI Phantom 3 Professional) blocking part of the photograph………………………..57

List of Boxes

Page number

Box 1.0 Scientific nomenclature of Z. noltii……………………………………….……………..12

Box 2.0 Future studies into the Z. noltii seedbank……………………………….…………..…56

Box 3.0 Potential developments for the use of a UAV in Z. noltii mapping…………………. 58

BI9999 PTY Report C1308785, A1317

6

List of Tables

Page number

Table 1 Physical factors that influence Z. noltii growth and seagrass distribution………..…10

Table 2 Review of previous forms of remote sensing that have been used to monitor and map global seagrass populations………………………………………………………………….14

Table 3 Range of data sources used to identify historical locations of Z. noltii found in South and West Wales……………………………………………………………………………………..15

Table 4 Geographical location of the study sites, historical and current record of Z. noltii presence and initial anthropogenic threats observed nearby (within a 500m radius) the Z. noltii meadows……………………………………………………………………………………….17

Table 5 Differences in the technical specifications between the three unmanned aerial vehicles (UAV) models used in this study……………………………………………………..…19

Table 6 Parameters used in the steps to generate orthomosaics and DEM’s from aerial data in Agisoft Photoscan Pro 1.2.0…………………………………………………………………….22

Table 7 Surface area of Z. noltii at the 10 study sites where seagrass was found to be present along the South and West Wales coastline……………………………………...……..27 Table 8 Differences in the time spent using a handheld GPS and light weight UAV to map the spatial and temporal distribution of Z. noltii meadows……………………………………...32 Table 9 Seagrass parameters recorded at 13 of the locations where Z. noltii was found to be present along the South and West Wales coastline…………………………………...………..41

Table 10 PCA results of 10 Z. noltii sites sampled along the South and West Wales coastline.................................................................................................................................44

Table 11 Report card for the seagrass condition for Z. noltii meadows in the South and West Wales region (March – July 2016). Zostera condition score and index were calculated using seagrass cover, shoot density and canopy height. Dark green indicates very good condition, green is good condition, yellow is moderate condition, orange medium poor condition and red very poor condition……………………………………………………………………..………46

Table 12 European Z. noltii meadows, where expansion of the extent has been recorded in previous literature…………………………………………………………………………….……..51

Table 13 Possible explanatory mechanisms for the expansion of Z. noltii at specific study sites (PE, AB, PW, PD and WW) from 2004……………………………………………………..53

Table 14 Estimates of Z.noltii seed density recorded in previous literature………………….55

Table 15 The decision rules used to classify Z. noltii beds that had previously been remotely sensed with SPOT multispectral images…………………………………………………………58

BI9999 PTY Report C1308785, A1317

7

1.0 Introduction

Seagrass are marine flowering plants (angiosperms) that form extensive beds

in shallow and sheltered coastal lagoons and estuaries, providing an essential

habitat for many coastal species (Short et al. 2011). In the UK two species of

seagrass are recognised; Zostera marina (eelgrass) and Zostera noltii (dwarf

eelgrass). Z. noltii (Figure 1) is an intertidal species rarely found past the low tide

mark (Stace 1997), with its distribution extending along the Atlantic coasts ranging

from south Mauritania (Cunha and Araujo 2009) to the southern fjords of Norway

(den Hartog 1970)(Figure 2). It is commonly found adjacent to saltmarsh

communities and relies on specific environmental conditions in order to grow and

survive (Table 1). Z. noltii is ecologically and economically important, acting as an

ecosystem engineer by attenuating waves (Paul and Amos 2011) and encouraging

the accretion of sediments. The roots and rhizomes of the plant help to stabilise the

substratum, improving water quality and thus creating a dynamic favourable habitat

that displays higher levels of biodiversity than other unvegetated areas (Bertelli and

Unsworth 2014). Polte (2005) displayed how Z. noltii meadows function as nursery

and extended juvenile habitats for epibenthos and provide a source of food to

wildfowl including Branta bernicla (Brent goose), Anas Penelope (Eusasian wigeon)

and Cymus olor (Mute Swan) (Davison and Hughes 1998). Furthermore, seagrass

meadows act as carbon sinks, accounting for 10-18% of the total oceanic carbon

burial despite only covering 0.1% of the world’s ocean floor (Mcleod et al. 2011).

Figure 1 Zostera noltii submerged in surface water located at Penclacwydd, Carmarthenshire. Source: authors own, 2016.

BI9999 PTY Report C1308785, A1317

8

Table 2 Physical factors that influence Z. noltii growth and seagrass distribution.

Environmental Factor Optimum Z. noltii Conditions

Depth and Light

availability

Like all plants Z. noltii requires light to photosynthesis. The intertidal species

is found at a range of depths from 0-10m (Short et al. 2011) and tends to

survive in shallower, high light intensities than Z. marina due to its small

rhizomes that limit the plant’s ability to store carbohydrates. Z. noltii has also

displayed signs that it is able to adapt to differences in water depth by

adjusting its growing behaviour. In shallow waters it produced dense but

short meadows, whereas longer leaves are often found at deeper sites.

Sediment Z. noltii grows on sandy to muddy substrata (Davison and Hughes 1998).

Unlike other Zostera spp, Z. noltii prefers finer-textured sediments that

provide higher fertility, allow rhizome elongation and have greater levels of

anoxia. Fine soft sediments are often found in sheltered, low energy

environments that remain partly submerged to stop the roots from drying out.

Temperature Z. noltii can survive at a range of temperatures from 7⁰C to 23⁰C and the

optimum temperature for growth and germination appears to be between

10⁰C -15⁰C (Davison and Hughes 1998). The average temperatures

observed in Wales (winter temperatures around 8⁰C and summer highs

around 17⁰C) are within this range therefore temperature is unlikely to be the

limiting factor of Z. noltii distribution.

Salinity Z. noltii is a euryhaline species (Paul 2011) that can tolerate high salinities

(15-30%) but also survive in brackish conditions down to salinities less than

5% (Charpentier et al. 2005). However Z. noltii generally prefers lower

salinity conditions for germination (optimum 1.0% salinity) and therefore

freshwater run off can be said to influence the distribution of Z. noltii

(Hootsmans et al. 1987)

Figure 2 Geographical locations of Z. noltii

(dwarf eelgrass) in Europe, possessing a

smaller range that other Zostera spp. Source:

Borum and Greve, 2004.

BI9999 PTY Report C1308785, A1317

9

1.1 Z. noltii status

Seagrass species are globally in decline (Green and Short 2003), significantly

reducing since the 1970s (Figure 3). Z. noltii is listed in the Least Concern category

of the International Union for the Conservation of Nature’s (IUCN) Red List of

Threatened Species, primarily due to its large range size, however Z. noltii has

displayed declining population trends (Short et al. 2010). Zostera communities in

Wales are listed as a UK Biodiversity Action Plan habitat (Maddock 2011), with

declines in the seagrass abundance thought to be caused by a variety of factors

including Zostera wasting disease, climate change (reducing the metabolism of the

plant) and eutrophication from anthropogenic threats (Jones and Unsworth 2016). It

is therefore important to map and understand the current distribution of Z. noltii

meadows, to help monitor and safeguard the future of this marine habitat. Seagrass

species are also considered indicators of water quality and health, due to their

susceptibility to changes in environmental factors (Pergent-Martini and Pergent

2000; Bhattacharya et al. 2003), with degraded habitats influencing their

effectiveness as ecosystem stabilisers (Uhrin and Townsend 2016). Therefore,

monitoring the condition of Z. noltii populations is vital.

1.1 Z. noltii monitoring

There exists limited detailed information about the current distribution of Z.

noltii in South and West Wales. Mapping of Zostera meadows is inherently difficult

due to factors such as unfirm substrata, cold temperatures and high winds/ground

Figure 3 Global decadal trends in seagrass extent, with sites either decreasing, increasing or displaying no change in the size of a meadow. Adapted from Waycott et al (2009).

BI9999 PTY Report C1308785, A1317

10

swell/wind swell causing changes in water clarity leading to poor visibility. Historical

records of Z. noltii are sparse and often incomplete with the volume, accuracy and

reliability of data taken from the marine ecosystem being significantly less than that

of terrestrial environments (Carr et al. 2003). The extent of the data is often limited,

only focusing on well-known locations and many of the records only display habitat

occurrence through point formats that do not represent the spatial and temporal

changes in the dynamic Z. noltii habitat. Many of the historical records of Z. noltii

were falsely identified as Z. angustifolia, the phenotypic variation of Z. marina. This

previous taxonomic confusion, in addition to differing nomenclature (Box 1.0), has

led to a severe lack of understanding and underestimate of Z. noltii populations in

Wales (Kay 1998). Comprehensive strategies for both the mapping and inventorying

changes in the Z. noltii habitats are needed; to clarify the current distribution of the

species and to derive the biophysical properties of the meadows (Armstrong 1993;

Koedsin et al. 2016) that have not been previously recorded. Predictive habitat

distribution models are also increasingly being used to map the species location

where there is a lack of resources (Grech and Coles 2010; Brown 2015). An

accurate geographical information systems (GIS) map of the Z. noltii meadows in

South and West Wales could therefore help in the development and to quantify

results of future models.

1.4 Unmanned aerial vehicles (UAVs) in Z. noltii monitoring

Remote sensing and manual delineation of marine ecosystems through

photointerpretation is becoming widely used by researchers, as it enables scientists

to map and assess fragile marine environments without disturbing them (McEvoy et

al. 2016). Previous studies have focused on obtaining data from larger remote-

sensing instruments that are not designed for detailed biological monitoring

(Anderson and Gaston 2013) (Table 2). However, developments in the use of UAV

Box 1.0 Scientific nomenclature of Z. noltii

For the purpose of this study dwarf eelgrass is referred to as Z. noltii to avoid

confusion with previous scientific literature. However, on the World register of

marine species Zostera noltii is not accepted and instead recorded as Z. noltei.

There is also current discussion as to whether dwarf eelgrass may in fact be of a

separate genus, NanoZostera noltii (Coyer et al. 2013).

BI9999 PTY Report C1308785, A1317

11

technology have meant that using remote sensing has become a more accessible

tool for smaller teams and projects, with such developments providing the platform

for UAVs to be used as effective mapping tools for seagrass meadows. Meadows

often have varying percentage covers, from extensive and continuous to

aggregations of patchy mounds, which are difficult to include in mapped polygon

data. The use of aerial data for remote sensing allows unvegetated parts of a

meadow to be mapped within a polygon, providing an accurate and comprehensive

map of the spatial distribution of an individual meadow compared to point and

transect based surveys (Lyons et al. 2011; Koedsin et al. 2016). The use of drone

technology also provides an alternative in areas where the water is too deep or

conditions too turbid for satellite remote sensing. However, some seagrass mapping

results revealed low accuracy in some of the biophysical measurements compared

to field observations, with sparse areas often being misclassified (Knudby and

Nordlund 2011). The use of neo-infared cameras on a UAV offers potential solutions

to misclassification (Barille et al. 2009). Assessing the ability of this form of remote

sensing to map intertidal Z. noltii populations will help to provide valuable data and

evidence in the development of novel drone-based approaches.

1.5 Aims and Objectives

This study describes an assessment of the distribution and status of Z. noltii in

South and West Wales.

The aims are to;

1) Identify changes in the physical location of Z. noltii populations compared

to historical distributions.

2) Trial photographic mapping at specific sites and assess the ability of a

drone as an effective Z. noltii mapping tool.

Outcomes of the study were to generate a series of vector data over a wide spatial

scale across South and West Wales and look at how the knowledge of this species

distribution can help in its conservation.

BI9999 PTY Report C1308785, A1317

12

Table 2 Review of previous forms of remote sensing that have been used to monitor and map global seagrass populations. Adapted from Hossain et al, (2015).

Aerial data platform Examples Published Studies

Hyperspectral Satellite Hyperion Lee et al. (2005), Pu and Bell (2013), Pu et al. (2012)

Multispectral Satellite (High spectral resolution)

IKONOS, Quickbird, WorldView2 Maeder at al. (2002), Mumby and Edwards (2002), Hochberg, Andrefouet and Tyler (2003), Dekker, Brando and Anstee (2005), Purkis and Riegl (2005), Wolter, Johnston and Niemi (2005), Fornes et al. (2006), Mishra et al. (2006), Habeeb et al. (2007), Karpouzli and Malthus (2007), Phinn et al. (2008), Sagawa et al. (2008), Yuan and Zhang (2008), Dingtian (2009), Dogan, Akyerk, and Belioglu (2009), Howari et al. (2009), Urbanski, Mazur and Janas (2009), Yang and Yang (2009), Amran (2010), Knudby and Nordlund (2011), Lyons, Phinn, and Roelfsema (2011), Liew and Chang (2012), Hamylton, Hagan, and Doak (2012), Phinn, Roelfsema, and Mumby, (2012), Baumstark et al. (2013), Borfecchia et al. (2013), Lee, Weidermann, and Arnone (2013), Roelfsema, Phinn et al. (2013), Tamondong et al. (2013), Roelfsema et al. (2014),

Multispectral Satellite (Medium spectral resolution)

Landsat MSS, Thematic Mapper, Enhanced Thematic Mapper+, ALI,

China Brazil Earth Rources Satellite, Satellite Pour I’Observation de la Terre, Advanced Land Observing Satellite, Advanced Spaceborne Thermal Emission and Reflection

Radiometer

Ackleson and Klemas (1987), Lennon and Luck (1989), Khan, Fadlallah, and Al-Hinai (1992), Armstrong (1993), Mumby et al. (1997), Ferguson and Korfmacher (1997), Andrefouet et al. (2001), Chauvard, Bouchon, and, Maniere (2001), Mumby and Edwards (2002), Purkis et al. (2002), Bouvet, Ferraris, and Andrefouet (2003), Call, Hardy, and Wallin (2003), Lunden and Gullstrom (2003), Schweizer, Armstrong and Posada (2005), Pasqualini et al. (2005), Gullstrom et al. (2006), Shapiro and Rohmann (2006), Vanderstraete, Goossens, and Ghabour (2006), Cho (2007), Lauer and Aswani (2008), Matarrese et al. (2008), Phinn et al. (2008), Wabnitz et al. (2008), Howari et al. (2009), Roelfsema et al. (2009), Yang and Yang (2009), Barille et al (2010), Knudby et al. (2010), Zhang (2011), Lyons, Phinn and Roelfsema (2012), Meyer and Pu (2012), Pu et al. (2012), Chen at al. (2013), Pu and Bell (2013), Roelfsema, Kovacs, et al. (2013), Torres-Pulliza et al. (2013), Wicaksono and Hafizt (2013), Knudby et al. (2014)

Multispectral Satellite (Low spectral resolution)

Moderate Resolution Imaging Spectrometer, Medium Resolution

Imaging Spectrometer

Kutser, Vahtmae, and Martin (2006), Saulquin et al. (2013), Petus et al. (2014)

Aircraft (Laser) Lidar Wang and Philpot (2007), Chust et al (2008), Tulldahl and Wikstrom (2012)

Aircraft (Hyperspectral) Compact Airborne Spectrographic Imager, HyMap

Mumby et al. (1997), Mumby and Edwards (2002), Habeeb et al. (2007), Peneva, Griffith, and Carter (2008), Phinn et al. (2008), Brando et al. (2009), Casal et al. (2013)

Aerial (Photograph) AP, HR digital camera Benton and Newman (1976), Larkum and West (1990), Orth, Moore, and Nowak (1990), Robins (1997), Chauvaud, Bouchon, and Maniere (1998), Kendrick, Eckersley, and Waler (1999), Kendrick et al (2000), Moore, Wilcox, and Orth (2000), Seddon, Connolly, and Edyvane (2000), Lathrop et al. (2001), Pasqualini et al (2001), Andrefouet et al. (2002), Cuevas-Jimenez, Ardisson and Condal (2002), Kendrick et al. (2002), Agostini et al (2003), Leriche et al. (2004), Meehan, Williams, and Watford (2005), Zharikov et al. (2005), Aswani and Lauer (2006), Hernandez-Cruz, Purkis, and Riegl (2006), Leriche et al. (2006), Habeeb et al. (2007), Holmes et al. (2007), Mount (2007), Mudoch et al. (2007), Reise and Kohlus (2008), Young et al (2008), Cunha and Santos (2009), Fletcher, Pulich, and Hardegree (2009), Andrade and Ferreira (2011), Cuttriss, Prince, and Castley (2013), Dolch, Duschbaum, and Reise (2013), Buttger, Nehls, and Stoddard (2014)

Boat (Acoustic) Side scan sonar, SDES, MBES, Acoustic Doppler current profiler,

Sediment Image Sonar

Hughes Clarke, Mayer and Wells (1996), Siccardi, Bozzano, and Bono (1997), Komatsu and Tatsukawa (1998), Pasqualini et al. (1998), McCarthy and Sabol (2000), Pasqualini et al. (2000), Sabol et al. (2002), Komatsu et al. (2003), Shono et al. (2004), Jordan et al. (2005), Riegl et al (2005), Siwabessy et al. (2006), Karpouzli and Malthus (2007), Ryan et al. (2007), Warren and Peterson (2007), Wilson et al. (2007), Lo Iacono et al. (2008), Sagawa et al. (2008), Lefebvre et al. (2009), Marsh and Brown (2009), Ahsan et al. (2010), Brooke, Creasey, and Sexton (2010), De Falco et al. (2010), Dekker et al. (2011), Descamp et al. (2011), Di Maida et al. (2011), Hamilton and Parnum (2011), Paul et al. (2011), Hasan, Ierodiaconou, and Laurenson (2012), Micallef et al. (2012), Sanchez-Carnero et al. (2012), Montefalcone et al. (2013), Munday, Moore, and Burczynski (2013)

Boat (Laser) FILLS Dierssen et al. (2003)

BI9999 PTY Report C1308785, A1317

13

2.0 Materials and methods

2.1 Study sites

The welsh coastline stretches 1680 miles (2740km)(WWF 2012), bordering

the Irish Sea and Bristol Channel. Z. noltii populations in Glamorgan,

Carmarthenshire and Pembrokeshire were assessed, and using all known historical

data of seagrass coverage in the region, sites were revisited to examine temporal

change. Historical records were obtained from a range of sources (summarised in

Table 3) and areas that have environmental parameters where Z. noltii has the

potential to grow and thrive. Those sites suggested in Brown’s (2015) potential

distribution model were also sampled. Locations where records of Z. noltii consisted

of leaves wash up on a beach were not used, as leaf litter can travel large distances

(McMahon et al. 2014) and “cast up” does not necessarily indicate the presence of a

nearby meadow (Kay 1998).

23 intertidal locations (Figure 4) along the South Wales coastline were

assessed during March-July 2016 at low tide. Unfortunately, Z. noltii in the Pembroke

River was unable to be included in the study. At each site, the presence or absence

of Z. noltii was recorded and habitat data was collected in the field. Details on the

anthropogenic impacts in the area were recorded (Table 4) to put the data in context,

and identify threats that might influence Z. noltii distribution (Jones and Unsworth

2016).

Table 3 Range of data sources used to identify historical locations of Z. noltii found

in South and West Wales.

Data Source Data type

NRW (formally Countryside

Council for Wales, CCW)

Reports, GIS layer, Point and Polygon shapefiles of Z. noltii presence,

Global Positioning System (GPS) latitude/longitude with 10m error.

Botanical Society of Britain

and Ireland

Z. noltii presence points only

(Kay Q, 1994) Report summarizing data collection from county records and other key

sources.

BI9999 PTY Report C1308785, A1317

14

Figure 4 Map of the seagrass study sites visited along the South and West Wales coastline. Orange indicates sites that have a historical record of Z. noltii and yellow indicates sites that have been predicted to have suitable conditions for Zostera growth (Brown 2015). Ruppia maritime historical records were also found at Aberthaw lagoon, in addition to displaying suitable Zostera growth conditions. Source: L. Pratt, ArcMap 10.2.2, 2016.

Legend

BI9999 PTY Report C1308785, A1317

15

Table 4 Geographical location of the study sites, historical and current record of Z. noltii presence and initial anthropogenic threats observed nearby (within a 500m radius) the Z. noltii meadows. Ranking of the anthropogenic threats is indicated by colour; red (extremely high), orange (high), yellow (adequate), green (low) and dark green (extremely low). Adapted from Jones (2014).

Site Site Code

Coordinates Date Visited Most recent Zostera historical record

Z. noltii presence in 2016

Industry Tourism Agriculture Catchment Population Anchoring Permanent Mooring

Ranking of threats

Jackson Bay

JB 51°23'26.04"N 3°15'48.36"W 9 May 2016 - Absent X X 2

Whitmore Bay

WB 51°23'18.14"N 3°16'25.71"W 9 May 2016 - Absent X X 2

Watch House Bay

WH 51°23'17.77"N 3°17'19.26"W 10 May 2016 - Absent X X 2

Dams Bay DB 51°22'55.73"N 3°20'08.62"W 20 May 2016 - Absent X 1

Aberthaw AT 51°22'54.17"N 3°23'04.31"W 12 April 2016 1998 Absent X X 2

Blackpill BP 51°35'40.97"N 3°59'21.41"W 10 April 2016 1985-1995

Absent X X X X 4

Norton NO 51°34'48.85"N 3°59'39.91"W 11 April 2016 1985-1995

Absent X X X 3

Oxwich Bay OB 51°33'33.88"N 4°09'18.32"W 6 May 2016 1974 Absent X X X 3

Broughton Bay

BB 51°37'21.13"N 4°16'50.31"W 6 April 2016 1994 Absent X X 2

Burry Pill BP 51°37'59.56"N 4°13'57.32"W 21 March 2016

2004 Present 0

Llanrhidian Sands

LS 51°38'01.79"N 4°11'21.24"W 24 June 2016 2013 Present X 1

Llanmorlais LL 51°38'36.87"N 4°07'33.01"W 23 April 2016 1916 Absent X X X 3

Penclawydd PE 51°39'18.66"N 4°08'21.50"W 23 March 2016

2013 Present X X X X 4

Ferryside FS 51°45'20.35"N 4°22'38.72"W 4 May 2016 1947 Absent X X 2

Angle Bay AB 51°40'46.64"N 5°02'43.59"W 12 March 2016

2013 Present X X X 3

Pwllcrochan PW 51°41'25.14"N 5°00'28.76"W 23 June 2016 2013 Present X X X X 4

Pembroke Dock

PD 51°41'47.83"N 4°56'45.01"W 6 June 2016 2013 Present X X X X X X 6

West Williamson

WW 51°43'12.86"N 4°51'50.51"W 8 July 2016 2013 Present X X X X 4

Garron Pill GP 51°44'03.31"N 4°52'47.97"W 10 June 2016 2013 Present X 1

Sprinkle Pill SP 51°45'37.52"N 4°53'57.43"W 22 May 2016 2013 Present X X 2

BI9999 PTY Report C1308785, A1317

16

2.2 Mapping of the intertidal Z. noltii meadow boundary

If Z. noltii was present and the meadow had a clear definitive boundary, it was

mapped by walking around the meadow edge with a handheld GPS unit (Garmin

GPSmap 62 and GPSmap 62stc) in tracking mode (accuracy error of approximately

1m). The data was then incorporated into the ArcGIS software by ESRI

(Environmental Systems Resource Institute), specifically ArcMap 10.2.2, for

geoprocessing using the British National Grid coordinate system. Single polygons

(Duggan-Edwards and Brazier 2015) were generated, from which the area of the

meadow was calculated and compared to the historical records. If any changes were

observed the percentage of regression and expansion was calculated (Telesca et al.

2015). However, if the meadow was extremely large (>100km2) and patchy, with an

inconsistent plant density (changing throughout the bed), transects were sampled

perpendicular to the shoreline, starting where Z. noltii was present. Observers

marked GPS waypoints every 50m and the presence/absence of Z. noltii along with

the percentage cover was recorded. Transects were 100m apart, enabling an overall

sampling grid to be created over a single bay and percentage cover was assessed

by placing a 50cmx50cm stainless steel quadrat consistently in front of the right foot

of the navigator, whilst walking along the area grid. Meadow boundaries were

determined by continuing along the transects until three successive points indicated

no presence of Z. noltii (Rasheed et al. 2003) and transects continued along the

shore until seagrass was no longer found.

2.3 Drone Based data collection

Mapping the Z. noltii meadows on foot was only feasible in areas with

relatively firm substrata, therefore this method could not be used in areas such as

Garron Pill, where the substrate is soft mud (Bunker 2008). In these inaccessible

areas a drone was used as a platform from which to collect photographs, that could

be stitched together to form a complete mosaic of a meadow. A DJIGlobal Phantom

3 Advanced UAV (Table 5) with a built-in Sony EXMOR1/2.3”, was used in its

autonomous flight mode to capture overlapping JPEG images (12 megapixels). It

was flown along a transect so that captured images could be incorporated into a

larger mosaic and was flown using the DJI GO app for iPhone 5S controlled by a

pilot and an observer. Prior to flight linear transects were plotted onsite, to avoid any

BI9999 PTY Report C1308785, A1317

17

obstacles in the field. Multiple missions were conducted at fixed heights, to ensure

the whole meadow was captured and to depict the optimum height at which Z. noltii

can be identified. To avoid sun glints effects, flights were preferentially carried out

early morning or late evening during cloudy weather, if compatible with low tides

(Jaud et al. 2016). Aerial images were also taken on a customized 3DR solo and

3DR Y6, where a similar method was used to assess the use of the different models

as environmental mapping tools. Aerial photographs were captured with a twin set of

modified Xiaomi Yi Action Cameras (both RGB and NIR/Blue filtered) and a RICOH

GR II consumer grade compact camera where both RAW and JPEG images were

recorded, to test the ability of the different cameras. The NIR/Blue filtered camera

was also used to assess the spectral properties of seagrass related to the reflective

part of the visible and infared spectrum that cannot be observed through

panchromatic photographs. The drones were flown using waypointed routes plotted

using Mission Planner and flown using pixhawk autopilot (Figure 5). They were

typically flown at a velocity of 5m/s to guarantee at least 50% overlap between the

successive images (Figure 6)

Table 5 Differences in the technical specifications between the three UAV models used in this study. Source of the model images: authors own, 2016.

DJIGlobal Phantom 3

Advanced UAV Customized 3DR Y6 Customized 3DR Solo

Weight (including battery and propellers) (g)

1280

2500

1800

Max Speed

16m/s (ATTI mode/no wind)

10m/s 55mph (89 km/h)

Max Flight Time (minutess)

23 10 20

BI9999 PTY Report C1308785, A1317

18

> 9

Figure 6 Example of overlapping images and computed camera positions recorded at A) Pwllcrochan and B) Garren Pill, C) fine- scale analysis at Angle Bay and D) lower altitude flight at Garren Pill, 2016. Higher levels of overlap were observed on the RICOH GR II consumer grade compact camera as oppose to the modified Xiaomi Yi Action Cameras (Appendix 1). Colours indicate the number of overlapping images with dark blue presenting the highest overlap (<9) often in the centre of the image and red, with the lowest overlap (1) often at the edge of the TIFFs. Source: L, Pratt, 2016, using PhotoScan 1.2.0.

Figure 5 Example flight path of 3DR Solo at Garren Pill, superimposed on orthomosaic generated on Photoscan 1.2.0. Waypointed routes were plotted using Mission Planner and flown using pixhawk autopilot. Source: L, Pratt, 2016.

9 8 7 6 5 4 3 2 1

A B

D C

BI9999 PTY Report C1308785, A1317

19

3D textured models were then generated from the digital images using the

image processing software Agisoft PhotoScan Professional Version 1.2.0 (AgiSoft

2015). The models were created off 3 major steps (Figure 7) (Casella et al. 2016;

Ventura et al. 2016) using specific parameters (Table 6) and exported as

georeferenced TIFFs (Tagged Image File Format) for further analysis. TIFFs were

Alignment of the photographs using structure from motion (SFM) to form a point cloud (Ullman 1979)

Polygonal mesh generated from dense multiview stereo-matching (Seitz et al. 2006)

The mesh becomes textured with the original photographs to form a 3D model

Figure 7 Steps used to create a three-dimensional model in Agisoft PhotoScan Professional 1.2.0 (Details in italics represent the specific algorithms used and the layer present in the images is displayed in bold). Source: L, Pratt, 2016.

BI9999 PTY Report C1308785, A1317

20

Table 6 Parameters used in the steps to generate orthomosaics and digital elevation models (DEM) from aerial data in Agisoft Photoscan Professional 1.2.0.

then imported as a raster image into ArcMap 10.2.2 where composite habitat maps

were generated by combining the different spectral bands. This enabled Z. noltii

boundaries to be analysed using image classification methods (Dugdale 2007;

Meyer 2008; Jackson et al. 2011) (Appendix 2) and compared to the GIS polygon,

created by GPS points recorded on foot. At study sites where a modified Xiaomi Yi

Action Camera (NIR/Blue filtered) was used, the Normalised Difference Vegetation

Index was calculated (Barille et al. 2009; Valle et al. 2015) by combining the

reflectance of the red and infared red colour bands (Tucker 1979), to assess the

quantity of green biomass. However, it was often difficult to distinguish between Z.

noltii and other vegetation including microphytobenthos and Ulva spp (Bunker 2008),

therefore further ground truthing was required.

2.4 Morphological measurements

Random point sampling in meadows with relatively firm substrata was used in

addition to the aerial images and walking the meadow boundary, to confirm the

present of Z. noltii and check the accuracy of the UAV data. It also enabled the

Align Image Dense Cloud

Accuracy Medium Quality Medium

Pair pre-selection Generic Depth filtering Aggressive

Key point limit 40,000

Tie point limit 4,000

Mesh Texture

Surface type Height field Mapping mode Generic

Source data Sparse cloud Blending mode Mosaic

Face count Medium (30,000) Texture size/count 4096 x 1

Interpolation Enabled

Point classes All

NDVI = NIR – R

NIR + R

NIR = Neo-infared band R = Red band

BI9999 PTY Report C1308785, A1317

21

health and condition of Z. noltii meadows to be studied and using ArcMap 10.2.2, 30

random waypoints were generated within the meadow boundary. This not only

avoided bias but was more time efficient, as the points were loaded and easily

located on the Garmin GPSmap 61/GPSmap 62stc. Each waypoint was sampled

with a 50cm by 50cm stainless steel quadrat where the standard Seagrass-Watch

parameters were recorded (McKenzie et al. 2003). This included Z. noltii

presence/absence, seagrass percentage cover, shoot density, average canopy

height, average canopy width, algae cover, epiphyte cover, substrate and any other

seagrass species or organisms present. All surveyors were trained according to the

same methodology to confirm consistency between results and multiple readings of

some parameters such as percentage cover were taken to ensure quality assurance

(Bunker 2008). Estimations of percentage cover recorded in the surveying were

based on the area covered by the rooted plants and followed previous guidelines for

the species (Mazik and Boyes 2009). When the percentage cover was estimated to

be less than 5%, the number of individual shoots was recorded and the replicate

data was then averaged to provide a representative mean percentage cover and

shoot density for the entire meadow. At least 95% of the quadrats were also

photographed to ensure standardisation, calibration of observers and to provide a

permanent record to be used in further tools such as the citizen science app,

‘SeagrassSpotter’ (www.seagrassspotter.org).

2.5 Seed coring

Seed coring was carried out to monitor seagrass resilience at sites where Z.

noltii was present and to assess the viability of the technique. At sites where Z. noltii

was absent, 30 cores were also taken to indicate any signs of colonisation and

potential future distributions. Furthermore, the process of coring (summarised in

Figure 8) removed any bias from the winter die off.

2.6 Data Analysis

All morphological values were reported as means (±SD) unless stated

otherwise. Prior to analysis all data was tested for normality using Shapiro-Wilk in R

(R Core Team 2013) to meet the assumptions of parametric tests. However, if the

data failed to conform to the normal distribution, alternatives to the parametric tests

were used. Simple statistical tests such as two-sample t-tests and one-way ANOVA

BI9999 PTY Report C1308785, A1317

22

were conducted to identify any differences in the mean Z. noltii values obtained.

Kruskal-Wallis was used as an alternative non-parametric form of ANOVA, to test the

difference in habitat characteristics across study sites. Pearson product-moment

correlation coefficients were carried out to assess correlations between variables

such as rate of expansion/regression, seed density and Zostera Condition Index

(ZCI). Both simple linear (SLR) and non-linear regression models were used to

identify and predict associations between variables, for example between current

and historical Z. noltii extent and NDVI and seagrass morphometric results.

Each morphometric value recorded measured different aspects of seagrass

status and in order to identify which aspects presented most variability a principal

component analysis (PCA) was conducted (Zar 1984; Johnson and Gage 1997;

Salita et al. 2003; Jones and Unsworth 2016). Principal components (PCs) with

eigenvalues greater than 200 were considered, with variables less than or equal to -

0.4 or greater than 0.4 being selected as dominant in each component. A general

linear model (GLM) was then conducted using the dominant variables from the PCA

and other environmental factors to determine the differences between the habitat

characteristics (Gaussian: identity). Furthermore, a Zostera condition index (ZCI)

was calculated using the dominant components, based on previous techniques used

to measure habitat health (McKenzie et al. 2012, 2014; Wilde 2016). All of the

Figure 8 The process of coring; 1) Cores taken randomly (alongside random point sampling) with a standard PVC seed corer (50mm diameter x 250mm long) and cap (Mckenzie et al. 2003), pushed approximately 10cm deep into the substratum; 2) Sediment placed in stainless steel sieve (1-2mm mesh); 3) Sieve sediment using a little water and 4) Analysis of core content, where the number of whole seeds and half seeds retained was recorded. Seeds were then returned to the sediment to reduce core-induced disturbance. Image sources: authors own, 2016.

1) 2) 3) 4)

BI9999 PTY Report C1308785, A1317

23

habitat characteristics were perceived to represent a healthier meadow when higher

(Krause-Jensen et al. 2004) and the ZCI score ranged from 0-1, with 1 being the

healthiest meadow.

BI9999 PTY Report C1308785, A1317

24

3.0 Results

3.1 Z. noltii extent

At 11 of the 21 study sites Z. noltii was found to be present, with seagrass

covering a total of 73.3 hectares along the South and West Wales coastline (Table

7). The Loughor estuary and Milford Haven Waterway harboured all the seagrass

populations found, with significantly more Z. noltii being located in Pembrokeshire

(83.4%) compared to Carmarthenshire (16.5%). Mapping using the transects

technique was conducted at the two largest study sites (AB and PE), although was

not deemed as effective as using the handheld GPS (Figure 9). Z. noltii populations

were found at the Burry Pill study site but due to the extremely sparse distribution

and small meadow size (<1m2), seagrass morphometric sampling did not take place.

No Z. noltii was found in Ferryside, Oxwich and Swansea Bay despite the historical

presence (Kay 1998). There was also an absence of Z. noltii in Glamorgan, despite

being predicted as a suitable habitat for Z. noltii to grow (Brown 2015).

The size of Z. noltii meadows ranged from small isolated patches only

0.159ha in size (LS1), to 40.7 ha of extensive seagrass (AB) (7.3±12.3). Z. noltii

populations at Llanrhidian Sands and Sprinkle Pill displayed signs of regression (-

16.2%yr-1=±6.0) although it was difficult to quantify any change due to the extremely

small meadow size. No detectable change was observed at Garren Pill (-0.008%yr-

1). Signs of expansion were observed at the remaining study sites (16.6%yr-1±15.2),

displaying an overall average expansion of 2.0%yr-1±20.7 (median = 1.8%yr-1) for Z.

noltii meadows in South and West Wales. Although there was no significant

difference between the mean historical Z. noltii area compared to the meadow area

calculated from this study (t (9) = 0.2382, p = 0.817), a significant positive correlation

was observed between the two variables (r = 0.786, t (15) = 4.917, p < 0.001). Using

a simple-linear regression model the current (2016) extent of Z. noltii meadows that

were unable to be mapped in this study, was predicted (Figure 10).

BI9999 PTY Report C1308785, A1317

25

Table 7 Surface area of Z. noltii at the 10 study sites where seagrass was found to be present along the South and West Wales coastline (The year displayed in italics indicates the actual year in which field sampling took place, if it differs from the citation). Historical data of the Pembroke River study site that could not be sampled in this study is stated, but not included in the overall mean % rate of change, μ (yr-1), observed.

Size of Z. noltii Meadow (Ha) Mean % rate of

change, μ (yr-1)

Study Site (Hodges and Howe 2007) 1996

(Mazik and Boyes 2009) 2004

(Bunker, 2008)

(CCW GIS layer, 2008)

(Howson 2012) 2009

(Nikitik 2015) 2009

(Nikitik 2015) 2010

(Crook, Unpublished) 2011

(Nikitik 2015) 2011

(Nikitik 2015) 2012

(Natural Resources Wales GIS layer, 2013)

(Nikitik 2015) 2013

(Nikitik 2015) 2014

Current size from this study

Llanrhidian Sands 1 (LS1)

- 1.05 - - - - - - - - 1.51 - - 0.16 -18.43

Llanrhidian Sands 2 (LS2)

- - - - - 6.60 - - - 0.68 -17.94

Llanrhidian Sands 3 (LS3)

- 4.41 - - - - - - - - 1.16 - - 0.88 - 7.41

Penclawydd (PE)

- 1.63 - - 11 - - - - - 7.37 - - 11.64

23.77

Angle (AB) 5.22 - 27.19 - - - - - - - 32.77 * recorded in

Duggan - Edwards and

Brazier, 2015.

- - 40.68 16.07

Pembroke River (PR)

- - - - - 89.9 92.6 - 91.7 88.9 97.3 94.1 93.7 N/A 0.04

Pwllcrochan (PW)

- - - - - - - - - - 1.79 - - 3.85 38.34

Pembroke Dock (PD)

- - - - - - - - - - 3.20 - - 3.49 3.01

West Williamson (WW)

- - - - - - - - - - 6.56 - - 6.92 1.82

Garrens Pill (GP)

- - - 4.00 - - - - - - 4.95 - - 4.65 -0.01

Sprinkle Pill (SP)

- - - - - - - - - - 0.87 - - 0.32 -20.90

Average 7.33±12.25

2.04±20.68

BI9999 PTY Report C1308785, A1317

26

Figure 9 Differences in the area of seagrass recorded by 1) using a handheld GPS and 2) using transects, to assess different forms of methodologies in monitoring larger Z. noltii meadows (A, Angle Bay and B, Penclacwydd). Methods using the handheld GPS were more accurate with 34.6% more Z. noltii recorded at Angle and 205% at Penclacwydd, respectively. However, changes in the spatial distribution within a meadow could not be monitored therefore further monitoring should look at incorporating both methods, with transects on a smaller scale. Source: L. Pratt, ArcMap 10.2.2, 2016.

¯Coordinate System: British National GridProjection: Transverse MercatorDatum: OSGB 1936False Easting: 400,000.0000False Northing: -100,000.0000Central Meridian: -2.0000Scale Factor: 0.9996Latitude Of Origin: 49.0000Units: Meter

Legend

Export_Output

F__cover

10

25

50

75

100

Sheet1$ Events

Export_Output

0 0.045 0.090.0225 Miles

¯Coordinate System: British National GridProjection: Transverse MercatorDatum: OSGB 1936False Easting: 400,000.0000False Northing: -100,000.0000Central Meridian: -2.0000Scale Factor: 0.9996Latitude Of Origin: 49.0000Units: Meter

Legend

Export_Output

F__cover

10

25

50

75

100

Sheet1$ Events

Polygon2

0 0.045 0.090.0225 Miles

Legend

= Z. noltii % cover = Z. noltii absence

= Meadow Boundary

!!

!!

!!

!!

!!

!!

!

!!

!

!

!! !

!

!

!

!!

!!

!

!

!

!!

!

!!

!!

!!

!!

!!

!

!!

!

!!

!!

!

!!

!

!

!!

!

!!

!!

!!

!!

!

!

!!

!!

!!

!!

!!

!

!

!! ! !

!!

!!

!!

!

!!!

!

!

!!

!!

!!

! !! !

!!

!!

!

!

!

! !!

!!

!! !

!!

!

!!

!!

!!

!!

Legend

Angle%CoverMap

Export_Output

F__cover

1

10

100

! Sheet1$ Events

Export_Output_12

Export_Output1

¯ ¯0 0.09 0.18 0.27 0.360.045

Miles

Coordinate System: WGS 1984 Web Mercator Auxiliary SphereProjection: Mercator Auxiliary SphereDatum: WGS 1984False Easting: 0.0000False Northing: 0.0000Central Meridian: 0.0000Standard Parallel 1: 0.0000Auxiliary Sphere Type: 0.0000Units: Meter

A

¯Coordinate System: British National GridProjection: Transverse MercatorDatum: OSGB 1936False Easting: 400,000.0000False Northing: -100,000.0000Central Meridian: -2.0000Scale Factor: 0.9996Latitude Of Origin: 49.0000Units: Meter

Legend

Export_Output

F__cover

10

25

50

75

100

Sheet1$ Events

Polygon2

0 0.045 0.090.0225 Miles

¯Coordinate System: British National GridProjection: Transverse MercatorDatum: OSGB 1936False Easting: 400,000.0000False Northing: -100,000.0000Central Meridian: -2.0000Scale Factor: 0.9996Latitude Of Origin: 49.0000Units: Meter

Legend

Export_Output

F__cover

10

25

50

75

100

Sheet1$ Events

Export_Output

0 0.045 0.090.0225 Miles

B

GPS Transects

BI9999 PTY Report C1308785, A1317

27

The spatial distribution of existing Z. noltii meadows along the South and

West Wales coastline is also changing (Figure 11 and 12), with the expansion of the

seagrass meadows taking place mainly in the lower intertidal zone. The upper shore

meadow boundaries were limited by the increased elevation levels (max 12m above

chart datum) and populations were observed to be expanding further towards areas

of Salt Marsh (PE). At Pwllcrochan, Z. noltii populations were also observed to be

colonising further sheltered areas along the coastline.

Figure 10 Simple linear regression (SLR) model that can be used to predict the current Z. noltii extent based on historical (<2016) distributions. Pembroke River Z. noltii populations are predicted to have a current area of 121.71 ha based on the model, although the regression model is not site specific.

y = 1.2812x + 1.6592R2 = 0.5916P < 0.001

0

5

10

15

20

25

30

35

40

45

50

0 5 10 15 20 25 30 35

Cu

rren

t Ex

ten

t (h

a)

Historical Extent (ha)

BI9999 PTY Report C1308785, A1317

28

Figure 11 Evolution of the spatial distribution of Z. noltii from 2013 to 2016, at sites showing signs of localised regression; A) Llanrhidian Sands site 1 (LS1), B) Llanrhidian Sands site 3 (LS3), C) Garren Pill (GP) and D) Sprinkle Pill (SP). Llanrhidian 2 (LS2) was not included in the figure, as no historical records of the sites spatial distribution were available. Areas in green represent to most recent NRW Zostera GIS layer (2013) and meadow boundaries displayed in pink are those generated from this study. Z. noltii was mapped on foot using a handheld GPS apart from GP, where a customized 3DR solo was used to map regions of Z. noltii on infirm substrata. Source: L. Pratt, ArcMap 10.2.2, 2016.

D)

C)

B)

A)

Historical (NRW, 2013) Distribution 2016 Distribution Changes in the spatial distribution

BI9999 PTY Report C1308785, A1317

29

D)

A)

B)

C)

E)

Historical (NRW, 2013) Distribution 2016 Distribution Changes in the spatial distribution

Figure 12 Evolution of the spatial distribution of Z. noltii from 2013 to 2016, at sites showing signs of localised expansion; A) Penclacwydd (PE), B) Angle Bay (AB), C) Pwllcrochan (PW), D) Pembroke Dock (PD) and E) West Williamson (WW). Areas in green represent to most recent NRW Zostera GIS layer (2013) and meadow boundaries displayed in pink are those generated from this study. Z. noltii was mapped on foot using a handheld GPS. Source: L. Pratt, ArcMap 10.2.2, 2016.

BI9999 PTY Report C1308785, A1317

30

3.2 Drone analysis

Drone data was available for 2 of the study sites; Garren Pill where the

substrata was too deep to access via foot and Pwllcrochan, where aerial imagery

could be assessed against a known meadow boundary. Finer scale aerial data was

also collected at Angle Bay and Garren Pill, to compare the use of UAV photographic

mapping at relatively low and high seagrass densities. The use of a UAV at

meadows differing in size and ability of each camera to map Z. noltii was also

assessed. Aerial data capture permissions were obtained for all 3 of these sites,

however could not be obtained for other locations including Pembroke River and

West Williamson. A high number of recreational boat users also limited drone work

at West Williamson and data image capture was prevented in the Burry Inlet due to

high wind exposures.

The DJI Phantom 3 Professional was the simplest out of the 3 models to

operate, however it is not designed for environmental mapping and the lack of

waypointed flights meant mapping larger areas of dense Z. noltii (at sites such as

AB) was extremely difficult. The ability to fly waypointed missions enabled by

pixhawk driven 3DR drones made this form of remote sensing more effective and

enabled extended areas of Z. noltii that were unable to be assessed by foot, to be

mapped. The quality of each of the cameras and the ability to measure Z. noltii

morphometric data differed with each model. At low altitudes the RICOH GR II RGB

camera produced the brightest texture with the clearest exposure, however the

pixels often became distorted at high focus levels (Figure 13). The large area

coverage and repeatability of the drone models used also displayed how this form of

remote-sensing can be time efficient to smaller teams often working against strong

tidal cycles (Table 8).

Table 8 Differences in the time spent using a handheld GPS and light weight UAV to map the spatial and temporal distribution of Z. noltii meadows.

Fieldwork Activity Time taken on Handheld GPS (minutes)

Fieldwork Activity Time taken on UAV (minutes)

Setting up GPS 5 Planning mission 15

Walking around meadow boundary

*Dependant on size 60 (small, e.g. PW) 120 (large e.g. AB)

UAV image acquisition flight

*Dependant on model ~20

Collection of seagrass morphometric samples

120 Collection of seagrass morphometric samples

120

Total 3 hours 35 minutes Total 2 hours 35 minutes

BI9999 PTY Report C1308785, A1317

31

Figure 13 Differences in the quality of the aerial data recorded at Garren Pill, July 2016, on a customized 3DR Solo, flown at 15m using; A) RICOH GR II consumer grade compact camera, B) modified Xiaomi Yi Action Camera (RGB) and C) modified Xiaomi Yi Action Camera (NIR/Blue filtered). Ground resolution altered with each camera from 3.96 mm/pix on the RICOH, 4.57mm/pix on the RGB Xiaomi Yi Action camera and 4.53 mm/pix on the NIR/Blue filtered Xiaomi Yi Action camera.

B

A

C

0 0.01 0.020.005 Miles

¯

0 0.0005 0.0010.00025 Miles

BI9999 PTY Report C1308785, A1317

32

RGB Aerial data captured from the drones, indicated meadow boundaries

generated from the handheld GPS often provided an overestimate to the extent of Z.

noltii present at each study site (Figure 14), not taking into account differences in

spatial distribution across a meadow. Visual interpretations of unsupervised (Figure

15) and supervised classifications also identified a similar problem, with large areas

of microphytobenthos and nearby marshland often being misclassified for Z. noltii.

Supervised classifications displayed higher levels of accuracy than unsupervised

analysis (Figure 16), with the density of Z. noltii also influencing the affectability of

the classification. Whilst dense areas of Z. noltii were frequently detected and easily

mapped, sparse populations (<20% coverage) were often not identified. This

highlighted the difficulty in using spectral classifications from 3-band imagery to

determine seagrass cover and manual classification of the Z. noltii meadows was

then implemented to achieve accurate meadow boundaries (Figure 17). The NDVI

results also supported this as little visible difference was observed between Z. noltii

populations and surrounding vegetation (Figure 18). However, the calculated NDVI

values did allow Z. noltii areas to be differentiated from other intertidal habitats,

displaying an average index >0.3 (Figure 17).

Figure 14 Differences in the area of Z. noltii at A) Pwllcrochan and B) Garron Pill, recorded using Garmin GPSmap 62 and forms of classification of the UAV data carried out on ArcMap 10.2.2.

0

0.5

1

1.5

2

2.5

3

3.5

4

4.5

5

GPS Unsupervised Supervised Manual

Are

a o

f Z.

no

ltii

(hec

tare

s)

Image Classification

Pwllcrochan

Garron Pill

BI9999 PTY Report C1308785, A1317

33

A)

Figure 15 Unsupervised ISO Cluster classification of true-colour mosaics at both A) Pwllcrochan (30m) and B) Garren Pill (50m). Average accuracy of the classifications was 53.3% (K = 0.33 observed at Pwllcrochan, K = 0.42 observed at Garron pill) (Appendix 3). Red boxes indicate areas that display the same pixel type as Z. noltii, providing an over estimate of meadow area. Source: L. Pratt, ArcMap 10.2.2, 2016.

A)

B)

BI9999 PTY Report C1308785, A1317

34

Figure 16 Supervised Maximum Likelihood image classification (MLC) of true-colour mosaics at both A) Pwllcrochan (30m) and B) Garren Pill (50m). Average accuracy of the classification was 63.3% (K=0.48 at Pwllcrochan and K = 50 observed at Garron Pill) (Appendix 4). Red boxes indicate areas that display the same pixel type as Z. noltii, providing an over estimate of meadow area. Source: L. Pratt, ArcMap 10.2.2, 2016.

BI9999 PTY Report C1308785, A1317

35

A

C

Coordinate System: WGS 1984 World MercatorProjection: MercatorDatum: WGS 1984False Easting: 0.0000False Northing: 0.0000Central Meridian: 0.0000Standard Parallel 1: 0.0000Units: Meter

¯0 0.03 0.060.015 Miles

Coordinate System: WGS 1984 World MercatorProjection: MercatorDatum: WGS 1984False Easting: 0.0000False Northing: 0.0000Central Meridian: 0.0000Standard Parallel 1: 0.0000Units: Meter

¯0 0.03 0.060.015 Miles

B

Figure 17 The process of manual classification used to assess Z. noltii boundaries at Pwllcrochan; (A) The texture generated from PhotoScan Professional Version 1.2.0, (B) The meadow boundary generated from drawing visual polygons round the dense Z. noltii observed in the aerial imagery using ArcMap 10.2.2, heavily relying on the knowledge of the surveyor to define the clumped seagrass distribution and (C) Comparison between the size of the meadow boundary recorded from the DJI Phantom 3 Professional (1.2 ha, displayed in blue) to the data recorded on the portable GPS unit (3.9ha, displayed in green) (69.2% reduction). Source: Source: L. Pratt, ArcMap 10.2.2, 2016.

Coordinate System: British National GridProjection: Transverse MercatorDatum: OSGB 1936False Easting: 400,000.0000False Northing: -100,000.0000Central Meridian: -2.0000Scale Factor: 0.9996Latitude Of Origin: 49.0000Units: Meter

¯

BI9999 PTY Report C1308785, A1317

36

Figure 18 Relative NDVI map generated from true-colour and NIR imagery at Garren Pill, taken using a modified Xiaomi Yi Action Cameral (July 2016). Source: L, Pratt, ArcMap 10.2.2, 2016.

BI9999 PTY Report C1308785, A1317

37

Figure 19 Differences in the NDVI (Means±95%Cl) calculated for the different spectral classifications of the intertidal aerial data. The Z. noltii NDVI was significantly higher (F (4) = 5.152, p<0.001) than that observed in the other classifications, meaning that seagrass species can be clearly identified through this form of image analysis.

-0.3

-0.2

-0.1

0

0.1

0.2

0.3

0.4

0.5

ND

VI

Z. noltii

Microphytobenthos

Surface Water

Mud

Ulva. Spp

BI9999 PTY Report C1308785, A1317

38

3.3 Seagrass morphometrics

There was a distinct difference between the habitat characteristics of the 10 Z.

noltii meadows studied (Figure 20). Shoot density ranged from 4.1 to 125.1 per

0.10m2 (62.3±61.3), and varied significantly across the sites (H (9) = 157.2711, p <

0.001), with the highest values recorded at West Williamson and Pembroke Dock

and lowest on Llanrhidian Sands in the Burry Inlet (Table 9). Seagrass canopy height

was also significant across the different sites (H (9) = 81.128, p < 0.001), ranging

from 43.4mm to 74.4mm (59.9±29.5), with the highest lengths found similarly at

Pembroke Dock and lowest at Llanrhidian Sands. Seagrass percentage coverage

ranged from 3.8% to 71.9% (33.3±31.4), with again the highest values distributed

amongst the Pembrokeshire study sites and the lowest values observed at

Llanrhidian Sands. There was a significant difference in the percentage cover

recorded across the sites (H (9) = 178.8939, p < 0.001) and the estimation of these

specific seagrass parameters by different surveyors had no influence on the data

collected (H (2) = 2.3678, p-value = 0.306). All 3 of the habitat characteristics

mentioned above displayed a further significant difference between the average

shoot density (t (297) = 11.9292, p < 0.001), canopy height (t (298) = 2.6946, p <

0.05) and percentage cover (t (298) = 13.0569, p < 0.001) observed at expanding Z.

noltii meadows and beds displaying signs of regression. Expanding meadows

therefore displayed a higher shoot density, percentage cover and canopy height. No

correlation was observed between the in morphometric values and spatial

heterogeneity at each study site (Appendix 5).

The sediment composition altered across sites, comprising of mud, mud/sand,

sandy mud and mud/rock. The dense mud substratum was the most common,

accounting for 37% of the 300 quadrats sampled. The sandy/mud sediment and the

mud/sand substrate were also common representing 30% and 20% respectively,

with mud/rock accounting for the remainder. Furthermore, the majority of study sites

(82%) were identified as pure Z. noltii meadows, however reduced Z. marina

populations were found to inhabit the depressions and channels at specific study

sites (PE and PW).

BI9999 PTY Report C1308785, A1317

39

Table 9 Seagrass parameters recorded at 10 of the locations where Z. noltii was found to be present along the South and West Wales coastline.

Site Z. noltii cover (%)

Shoot density (per/0.2m2)

Leaf Length (mm)

Leaf width (mm)

Epiphyte cover (%)

Algae cover (%)

Llanrhidian Sands (Bed 1)

6.1±6.1 15.7±17.7 43.4±35.5 1.12±0.90 0.01±0.01 0.90±1.85

Llanrhidian Sands (Bed 2)

6.5±5.6 15.9±14.8 59.8±46.7 1.10±0.88 0.00±0.01 0.37±0.89

Llanrhidian Sands (Bed 3)

4.1±3.0 6.0±7.9 67.1±40.7 1.40±0.81 0.00±0.00 0.40±1.13

Penclawydd 21.5±15.6 46.4±24.7 47.7±11.8 0.95±0.30 0.00±0.00 0.00±0.00

Angle 30.5±24.5 64.0±41.4 46.4±17.5 0.94±0.37 0.01±0.04 0.10±0.40

Pwllcrochan 57.815±30.7 116.5±76.6 67.1±24.6 1.15±0.58 0.00±0.00 1.20±2.47

Pembroke Dock (West Llanion Pill)

66.1±30.0 116.8±66.6 74.4±16.8 1.56±0.52 0.00±0.00 0.53±2.22

West Williamson

71.9±23.0 125.1±61.7 79.4±22.9 1.43±0.51 0.00±0.00 4.03±7.32

Garrens Pill 44.8± 23.5 75.5± 31.0 66.7±14.3 1.23±0.44 0.01±0.01 0±0

Sprinkle Pill 24.3±18.8 41.2±43.9 47.1±14.6 1.03±0.18 0.0±0.00 0±0

Average 33.3±31.4 62.3±61.2 59.9±29.5 1.19±0.62 0.00±0.01 0.75±2.86

BI9999 PTY Report C1308785, A1317

40

Figure 20 Mean (±95%Cl) Z. noltii cover, shoot density, canopy height, canopy width, epiphyte cover and algae cover across all 10 study sites. Green indicates values that were recorded in a Z. noltii meadow that displayed signs of expansion, yellow no detectable change in the meadow size and red, a Z. noltii meadow with observed regression correlations in the area.

0

10

20

30

40

50

60

70

80

90

L1 L2 L3 PE AB PW PD WW GP SP

Z. n

olt

iiC

ove

r (%

)

0

20

40

60

80

100

120

140

160

L1 L2 L3 PE AB PW PD WW GP SP

Sho

ot

Den

sity

(p

er/0

.25

m2 )

0

10

20

30

40

50

60

70

80

90

100

L1 L2 L3 PE AB PW PD WW GP SP

Can

op

y H

eigh

t (m

m)

0

0.2

0.4

0.6

0.8

1

1.2

1.4

1.6

1.8

2

L1 L2 L3 PE AB PW PD WW GP SP

Can

op

y W

idth

(m

m)

0

0.005

0.01

0.015

0.02

0.025

0.03

L1 L2 L3 PE AB PW PD WW GP SP

Epip

hyt

e C

ove

r (%

)

0

1

2

3

4

5

6

7

8

L1 L2 L3 PE AB PW PD WW GP SP

Alg

ae C

ove

r (%

)

BI9999 PTY Report C1308785, A1317

41

3.4 Historical change in Z. noltii morphometrics

For 2 out of the 10 study sites historical changes over time of selected habitat

characteristics can be studied. There was a significant difference in the mean Z.

noltii cover (F (7) = 5.727, p < 0.05) and canopy height (F (5) =8.111, p < 0.05)

recorded from 1996 to 2013 at Angle Bay (Figure 21). However, no clear correlation

was observed in either of the seagrass morphometric values despite the increasing

aerial extent of the study site, mostly likely due to differences in methodologies.

Seasonality may also have influenced any correlations as differences in the

coverage of Z. noltii between April and June 2016 displayed how there was a

significant increase in the Z. noltii cover within these 3 months (t (58) = -4.3269, p <

0.05) to give a mean of 53.8±16.5.

Coverage data taken from the second Llanrhidian Sands site also displays an

increase in seagrass cover from 5.7±6.8 in 2011 to 6.5±5.6 in 2016, although there

was no significant difference between the means (t (129) = -0.5825, p = 0.5612) and

seasonality is likely to have had a similar effect on the results. However, there was a

significant increase in the shoot density per/0.25m2 at Llanrhidian 2 (t (129) = -2.945,

p < 0.05) despite the decrease in Z. noltii extent. The lack of extensive Z. noltii

surveying and accessible data meant that any further data analysis with historical

seagrass morphometrics was heavily limited.

Figure 21 Changes in the average Z. noltii cover (Means±95CI) and mean canopy height (mm) observed at Angle Bay from 1996 to 2000. Blue indicates historical values obtained from Hodges and Howe, 2007, yellow, Bunker, 2008, and green, Duggan-Edwards and Brazier, 2015, with the shoot density values obtained from the same data source. Red indicated values obtained from his study.

0

10

20

30

40

50

60

70

80

90

0

10

20

30

40

50

60

70

1996 1997 1998 1999 2000 2008 2013 2016

Can

op

y H

eigh

t (m

m)

Z. n

olt

iico

ver

(%)

Year

Series2

Poly. (Series2)

Canopy Height

Projected changes in

canopy height

BI9999 PTY Report C1308785, A1317

42

3.5 Multivariate analysis of morphometric variables

From my morphometric measurements, in order to determine which

characteristics contributed most to the variability within and between the different

study sites, a principle component analysis (PCA) was conducted. Two PCAs had

eigenvalues greater than 200 and together accounted for 96.5% of the variation

(Table 10). The first principal component (PC1) was the most important, having an

epigenetic value of 4720 and contributing to 84.2% of the variance. The component

was dominated by average Z. noltii percentage cover and shoot density, with the

axis suggested to be loosely representative of abundance, where an increase in PC1

represents an overall increase in the Z. noltii density (Figure 22). PC2 had a smaller

eigenvalue of 692 and contributed to 12.3% of the variation amongst the data, with

average canopy height being the only dominant variable.

Table 10 PCA results of 10 Z. noltii sites sampled along the South and West Wales coastline (Coefficients in bold represent the dominant variable in each principal component). For principal component scores see Appendix 6.

Principal Component 1 (PC1)

Principal Component 2 (PC2)

Summary Values

Eigenvalues 4720 692

% Variation 84.2 12.3

Cumulative % Variation 84.2 96.5

Seagrass Morphometrics

Average Z. noltii percentage cover

0.418* -0.020

Shoot Density 0.883* 0.245

Average Canopy Height (mm) 0.214 -0.969*

Average Canopy Width (mm) 0.003 -0.015

Epiphyte percentage cover 0.000 0.000

Algae percentage cover 0.005 -0.018

BI9999 PTY Report C1308785, A1317

43

From the results of the PCA, a general linear model (GLM) was conducted to

study the difference between the dominant components at each of the sites. The

influence of various factors including seasonality, sediment type and elevation were

also accounted for. As expected Z. noltii cover had a significant influence over the

shoot density (p < 0.001) and canopy height (p < 0.001) recorded (Appendix 7).

Shoot density however, has no influence over the canopy height observed (p =

0.298). The sediment type had a significant influence on Z. noltii cover (p < 0.05,

p<0.05, p<0.001, respectively) whilst elevation (p = 0.915, p = 0.523, p = 0.542;

GLM) and seasonality (p = 0.989, p = 0.611, p = 0.527; GLM) did not significantly

influence the data.

3.6 Zostera Condition Index

The 3 dominant components from the PCA, were used to indicate habitat

health and Zostera condition. Each study site was scored and ranked (Table 11) on

their respective condition, with the ranking suggesting that West Williamson is the

healthiest Z. noltii meadow sampled with a score of 11. Study sites at Llanrhidian

Sands had the poorest result with a seagrass health scores no greater than 6,

Figure 22 Principal component analysis (PCA) of seagrass characteristics from 10 different study sites, spread along the South and South West Wales coastline.

BI9999 PTY Report C1308785, A1317

44

Table 11 Report card for the seagrass condition for Z. noltii meadows in the South and West Wales region (March – July 2016). Zostera condition score and index (ZCI) were calculated using seagrass cover, shoot density and canopy height. Dark green indicates very good condition, green is good condition, yellow is moderate condition, orange medium poor condition and red very poor condition.

Site Z. noltii cover (%)

Shoot Density (per/0.2m2)

Canopy Height (mm)

Zostera Condition Score (ZCS)

Zostera Condition Index (ZCI)

L1 1 1 3 5 0.33

L2 1 1 3 5 0.33

L3 1 1 4 6 0.33

PE 2 2 3 7 0.47

AB 2 2 3 7 0.47

PW 3 3 4 10 0.67

PD 3 3 4 10 0.67

WW 3 4 4 11 0.73

GP 2 2 4 10 0.67

SP 2 2 3 7 0.47

matching the results of the multivariate analysis. A Pearson product-moment

correlation coefficient was then used to assess the relationship between ZCI and

change in the area extent of the Z. noltii meadows. A positive correlation was

displayed between the mean % rate of change (yr-1) (Appendix 8) and ZCI although

it was not significant (r = 0.2390563, df = 8, p-value = 0.506).

Z. noltii cover (%) Rank

>80 5

60 - <80 4

40 - <60 3

20 - <40 2

>20 1

Shoot Density (per/0.2m2)

Rank

>160 5

120 - <160 4

80 - <120 3

40 - <80 2

<40 1

Canopy Height (mm) Rank

>80 5

60 - <80 4

40 - <60 3

20 - <40 2

<20 1

BI9999 PTY Report C1308785, A1317

45

3.7 Z. noltii morphometrics vs NDVI

There was a positive relationship between NDVI and dominant Z. noltii habitat

characteristics (Figure 23), proving consistent with previous analysis (Valle et al.

2015). Significance in the correlation was observed between NDVI and Cover (r =

0.471, df = 57, p < 0.001) and NDVI and shoot density (r = 0.383, df = 57, p < 0.05)

through Pearson Product Moment coefficients, indicating the potential of NDVI to

monitor habitat health. However, the relationship between canopy height and NDVI

displayed no significance (r = 0.222, df = 57, p = 0.0912), indicating the limitations of

aerial data to monitor habitat health characteristics on a finer scale.

3.8 Core Analysis

Seeds (half and full) were found at 9 of the study sites, with the total number

of whole seeds found per site ranging from 0 to 3 (0.8±1.2) (Figure 24). The average

seed density per m2 for the Z. noltii meadows along the South and West Wales

coastline was 1.4±2.1 and there were significantly more seeds obtained at sites

where Z. noltii was known to be present (t (20) = 3.2042, p < 0.05). A Pearson

product-moment correlation coefficient also indicated there was a significant positive

correlation between average seed density per m2 and ZCI (r = 0.773, df = 20, p <

0.05) (Appendix 10), and seed density and mean % rate of change, μ (yr-1) (r =

0.705, df = 20, p < 0.001) (Appendix 11). Despite previous studies (Harrison 1993),

no significant correlation was observed between seed density and elevation of the Z.

noltii meadow (Appendix 12), most likely due to the limited number of seeds

recorded.

BI9999 PTY Report C1308785, A1317

46

Figure 23 Non-linear regression model estimating the relationship between the Normalised Difference Vegetation Index (NDVI) and dominant Z. noltii habitat characteristics; A) NDVI and cover, B) NDVI and shoot density and C) NDVI and canopy height (Appendix 9). Red represents data obtained from fine scale aerial imaging at Angle Bay and blue from images at Garren Pill. Data matches relationship previously calculated by Valle et al (2015).

A

B

C

y = -383.92x2 + 178.62x + 10.83R² = 0.2816P < 0.001

0

10

20

30

40

50

60

70

80

90

-0.05 0 0.05 0.1 0.15 0.2 0.25 0.3 0.35

Z. n

olt

iiC

ove

r (%

)

NDVI

y = -4399.34x2 + 949.54x + 14.24R² = 0.2528P < 0.001

-50

0

50

100

150

200

250

300

350

400

-0.05 0 0.05 0.1 0.15 0.2 0.25 0.3 0.35

Sho

ot

De

nsi

ty (

pe

r/0

,25

m2 )

NDVI

y = -824.41x2 + 166.44x + 46.97R² = 0.0686p = 0.051

0

20

40

60

80

100

120

140

-0.05 0 0.05 0.1 0.15 0.2 0.25 0.3 0.35

Can

op

y H

eig

ht

(mm

)

NDVI

Garron Pill Angle Bay

BI9999 PTY Report C1308785, A1317

47

3.9 Field observations

At all the study sites, personal observations recorded by the surveyors

indicated the risks to Z. noltii in the area. A number of the sites assessed had issues

with disturbance, with physical damage inflicted by commercial cockling activities

and bait digging (Figure 25). 6 out of the 10 study sites where Z. noltii was present

had boat anchoring or moorings present or in close proximity to the seagrass.

Elevated nutrient levels were also observed at one of the sites through increased

opportunistic macroalgae (Figure 26), effecting the water quality and light availability

to Z. noltii at the site. Patches of Z. noltii bleaching were also noted at Angle Bay,

although the causality was unknown (Figure 27).

Figure 24 Mean (±95%Cl) Z. noltii seed density (m2) at both sites where seagrass was present (green) and sites where Zostera spp were absent (red).

0

2

4

6

8

10

12

14

JB

WB

WH

DB

AT

BP

NO

OB

BB

BP

LS1

LS2

LS3 LL PE FS AB

PW PD

WW GP SP

Mea

n N

um

ber

of

Seed

s (m

2 )

Study Sites

BI9999 PTY Report C1308785, A1317

48

Figure 26 Opportunistic macroalgae observed at West Williamson, from extensive algae cover of the Z. noltii bed. Source: authors own, 2016

Figure 25 Area of intensive bait digging to the north of the intertidal Z. noltii bed in Angle Bay. Source: authors own, 2016.

Figure 27 Observed Z. noltii bleaching at Angle Bay, where the causality is unknown. Source: authors own, 2016.

BI9999 PTY Report C1308785, A1317

49

4.0 Discussion

Seagrass ecosystems are dynamic and ecologically important habitats that

are in need of monitoring (Borum and Greve 2004). This study, is the first

quantitative assessment of both the distribution and health of Z. noltii populations

along the South and West Wales coastline. Although study sites such as the

Pembroke dock could not be included, an overall expansion of Z. noltii in the Milford

Haven Waterway and Loughor estuary was observed. By revisiting historical sites,

effort has been made to distinguish the difference between sites where Z. noltii is not

present and sites where the data does not exist, with certain historical locations

displaying a seagrass loss. The evaluation of a UAV as a seagrass mapping tool

also highlights how this form of remote sensing can be used as an effective, novel

method to mapping intertidal marine habitats.

4.1 Seagrass status and extent

Z. noltii populations in South and West Wales are currently increasing in

spatial extent by 2.1% per year (1.54 ha) over a 20-year period, with site specific

regression and expansion patterns being observed. This finding is contrary to the

large scale global patterns of decline (Waycott et al. 2009). Although current

meadow expansion is clear, this study found 5 locations where seagrass has

previously been recorded but now absent. This highlighted extensive loss of Z. noltii

communities along the South Wales coastline despite the evidence of recent

increases. Limited baseline information from these sites prevented the ability of this

study to examine how much seagrass has been historically lost.

An overall increase in the mean rate of change (μ,yr-1) indicates a significant

expansion between the historical and current Z. noltii distributions, with the highest

increase observed at the largest meadows. Similar increases in the size of intertidal

Z. noltii meadows have been reported in other areas of western Europe (Table 12).

Defining causality for these changes was beyond the scope of the present study

however we provide a series of possible explanations for this, one of which is

improving water quality. The widespread decline in heavy industry may also be a

contributing factor (Table 13).

BI9999 PTY Report C1308785, A1317

50

Table 12 Only European Z. noltii meadows, where expansion of the extent has been recorded in previous literature. All other European meadows in the literature displayed signs of regression.

Location Source Most recent Z. noltii record

Total Surface Area (ha)

Mean % rate of change (yr-1)

Bourgneuf Bay, France (Barille et al. 2009) 2009 586 13.0

De Keeg-De Ans, Dutch Wadden Sea (Philippart and Dijkema 1995)

1987/1988 0.23 3.6

North Wadden Sea (Reise and Kohlus 2008)

2006 10020 18.0

Global Average 3535.41 11.53

Study Results 73.27 2.09±20.78

The upper limits of the Z. noltii meadows (on the shore) remained almost

unchanged, likely due to a threshold in which the negative effects of desiccation

cannot be counteracted (Barille et al. 2009). Expanding meadows displayed a higher

habitat health with increased Z. noltii coverage, shoot density and canopy height.

Elevated seed densities were also recorded at expanding meadows suggesting sites

have a significantly higher resilience and therefore more likely to survive against

potential threats. The observed gradual linear increase and good habitat condition

indicate that the major Sea Empress oil spill of 1996 (Edwards and White 1998) did

not significantly affect the long-term health and abundance of seagrass; confirmed

by field observations (Moore 2006) and the spatial extent of the Z. noltii meadow.

Study sites such as Llanrhidian Sands and Sprinkle Pill displayed signs of

regression most likely caused by changes in the substratum, from muddy sand in

which seagrass can thrive in to areas of rippled sand in which Zostera is unable to

grow (Mazik and Boyes 2009; Dolch and Reise 2010). Anthropogenic impacts

including activities such as intensive cockling in these small meadows is most likely

to have disrupted the growth of Z. noltii. Reduced coverage, shoot density and

canopy height was observed at these sites and repeated disturbance to the canopy

is likely to have reduced the seed output and deplete the seedbank for future

distributions (Unsworth et al. 2015). Vehicle’s observed driving across the seagrass

have led to further damage (Figure 27). Expansion of the Z. noltii extent was also

potentially limited at the West Williamson study site, due to increase in Ulva spp

BI9999 PTY Report C1308785, A1317

51

Table 13 Possible explanatory mechanisms for the expansion of Z. noltii at specific study sites (PE, AB, PW, PD and WW) from 1996. Adapted from Barille et al (2009).

Expansion Mechanism

Possible Outcomes Observed Effects

Increased Light Availability

a) Tidal emersion patterns change, due to local accretion

b) Turbidity of the water decreases. c) Reduced nutrient concentrations will

limit periphyton on blades and increase light availability.

a) Positive accretion observed by permanent cocklers at sites such as Penclacywdd. For sites that are expanding downwards such as Angle Bay it would justify the expansion, by improving the light conditions in the mid-intertidal zone as Z. noltii photosynthesise mainly at low tide.

b) No data available. c) Nutrient levels were obtained for the Milford Haven Waterway and displayed high levels associated to the sample areas. However, a slight decrease

in the dissolved inorganic nitrate (DIN) and an overall increase in the dissolved inorganic phosphorus (DIP) has been noted in the last 14 years (Natural Resources Wales 2014). Regression analysis from the 5 largest sewage treatment works (STWs) in the Milford Haven area also indicate a downward trend, with only ~8% of the DIN load attributing from continuous point sources such as STW (Natural Resources Wales 2016) and that the water quality in the study area is in fact increasing.

Increase in temperature

a) Increases the rate of photosynthesis b) Shift in the Z. noltii distribution

a) Sea surface temperatures are increasing (0.2⁰C/decade)(IPCC 2007), which increases the photosynthetic efficiency of PSII (Collier and Waycott

2014). Elevated temperatures will also increase the respiration rate, potentially adding an elevated metabolic burden (Marsh et al. 1986). Studies into the effects of higher temperatures on Z. noltii populations in Wales are limited and no effects were visibly observed. However, the relationship

needs long term monitoring as higher temperatures could rise above the Z. noltii tolerance (approximately 38⁰C, sampled from populations in Ria

Formosa), causing an increase in the synthesis of heat-shock proteins (HSP) (Massa et al. 2008) and negatively impact the plant. b) Global Z. noltii distributions are shifting northwards (expected to move more than 888km closer to the poles by the end of the 21st century) due to

increased seawater temperatures (Valle et al. 2014). Although this is unlikely to influence the current distribution of Z. noltii in Wales could have a long-term impact on seagrass species in the UK.

Increase in Salinity

a) Allow rapid colonisation of Z. noltii in new areas

b) No data available

Sea level rise (SLR)

a) Increasing the intertidal area available for Z. noltii growth.

a) No data available

Sediment becoming suitable

a) Decreased density of lugworms (Arenicola marina) that reduces the impact of bioturbation.

b) Change in the content and firmness of the sediment.

a) Little to none A. marina were observed in the field. Sites where more A. marina pits were observed such as Angle Bay, displayed reduced coverage, shoot density and canopy height. Lack of historical data and consistent seagrass monitoring in South and West Wales meant changes in A. marina distribution could not be analysed.

b) Further scientific analysis is required to detect a historical change. The results of the GLM do indicate that sediment influences the habitat characteristics and previous studies reveal how the Z. noltii coverage and extent was positively related to the clay content of the sediment (Philippart and Dijkema 1995). Clay layers provide a firm rooting for the plant and prevent the settlement of lugworms (Reise and Kohlus 2008).

Reduction in relative exposure index

a) Increased shelter provides a more suitable area for Z. noltii to grow and survive

a) Recent relative exposure index calculations reveal that sites in the Milford Haven Waterway have the lowest exposure in Wales (Brown 2015), supporting the expansion of Z. noltii meadows. Extensive coastal development in areas also increase local accretion making areas more suitable for Z. noltii growth.

Decrease in extreme weather events

a) Reduced drought effects will limit the impact on Zostera biomass and dynamics of species such as H. ulvae in the seagrass ecosystem.

b) Decrease in storms, reduces wave energy and suspended sediment.

a) Increased rainfall and higher frequency of heat-waves has been recently recorded in the UK, opposing this theory. b) No evidence for the decrease in the frequency and intensity of storms in the UK. However literature indicated that fewer mid-latitude storms

averaged over each hemisphere, with a poleward shift in the track and lower central pressure regions (IPCC 2007).

BI9999 PTY Report C1308785, A1317

52

cover most likely caused by DIN entering the waterway from nearby agricultural land

(Natural Resources Wales 2016). This reduces the seagrass light availability and

ability to photosynthesise. Similar small-scale declines have also been observed in

the Dutch Wadden Sea (OSPAR 2009), due to increased density of lugworms

resulting in younger shoots becoming smothered with sediment and intensive

grazing of Brenta bernicla.

Patchiness and further clumped distribution of Z. noltii was also mainly

observed at sites declining in size or at the boundary of expanding meadows (Figure

29). Fragmentation can therefore be used as an identification of a disturbed meadow

that is either in recovery or regressing after impact (Reed and Hovel 2006), leading

to reduced ecosystem connectivity.

Figure 2814 Tractor marks observed at Zostera meadow in Porthdinllaen, North

Wales. Source: Benjamin Jones, 2016.

BI9999 PTY Report C1308785, A1317

53

12 study sites lacked Z. noltii seeds and this is most likely due to the Zostera

seeds being heavy and negatively buoyant, therefore estimates of seed dispersal

suggest that the seeds only have the ability to move a few metres (Orth et al. 1994).

However, other studies consider long distance dispersal (occurring over 10s of

kilometres) as bare seeds, detached spathes and whole flowering shoots (Loques et

al. 1988) and that the absence of Z. noltii seeds at study sites could instead be due

to low viability and successful germination levels within the seedbank. Z. noltii seed

densities calculated within this study were much lower than those recorded in the

literature (Table 14). This is most likely due to the effects of seasonality as Z. noltii

forms new seeds at the end of August (Hootsmans et al. 1987), potentially reducing

seed densities in the seedbank from June to July, when a high majority of the

sampling took place. The present studies are the first to examine Z. noltii seeds in

the UK reflecting the limited understanding of seagrass reproduction in the region.

Further research into changes into Z. noltii reproduction and seedbanks are needed

to help understand the resilience of these meadows (Box 2.0).

Table 14 Estimates of Z.noltii seed density recorded in previous literature. Adapted from Zipperle et al. (2009).

Location Source Year of fieldwork Estimated Seed density in sediment (m-2)

Konigshafen Bay, Island of Sylt, German Wadden Sea

(Zipperle et al. 2009) 2004 487.5±269.4

2005 367.3±95.5

Zandkreek, The Netherlands

(Harrison 1993) 1989-1990 157±45.8

(Hootsmans et al. 1987) 1983 0-150

Ria Formosa Lagoon, Portugal

(Alexandre et al. 2006) 2003 312.1±66

Global Average 279.78±119.2

Study Average 1.4±2.1

BI9999 PTY Report C1308785, A1317

54

4.2 Affectability of drones as an alternative seagrass monitoring platform

The drone acted as an effective mapping tool, enabling wider spatial and

temporal distributions of the dynamic seagrass ecosystem to be recorded. Many of

the advantages to using a drone were addressed in section 1.4, with the time taken

to capture the aerial data significantly lower than that when using other methods due

to the facility to rapidly deploy and acquire photographs (Lomax et al. 2005)

(previously displayed in Table 8). This is key in image acquisition, with cloudy

conditions typical of the UK climate, often limiting optical remote sensing techniques

in previous studies. Instead the UAV can be kept in a state of readiness and only be

deployed in suitable atmospheric conditions, reducing the time taken to complete Z.

noltii surveys accurately.

There are limitations to using a UAV for seagrass mapping, such as the

lightweight nature making it more susceptible to wind and turbulence in comparison

to conventional aerial photography, leading to potential distortion of the aerial data

(Peterson et al. 2003) (Figure 29). This posed a significant problem at exposed study

sites in the Loughor Estuary, however other forms of remote sensing such as using

kites as an aerial platform are being investigated in these areas (Duffy and Anderson

2016). The operator skills may have also influenced the altitude and flight trajectory,

with a more experienced pilots producing more comprehensive and consistent

images compared to novel UAV operators (Dugdale 2007). Furthermore, the large

Box 2.0 Future studies into the Z. noltii seedbank

The expansion of a Z. noltii meadow in theory should cause an increase in the viable seed density. Different type of seeds in the age structured seed bank could be used to indicate the historical distribution of a seagrass meadow, based on the number of long-term seeds (<5yr) as oppose to the transient seeds (persisting for <1 year) (Thompson et al). Furthermore, the strong negative impacts of heat-stress on the seedbank can be used to determine how environmental factors can influence the distribution of seagrass.

However, literature displays that only 12% of the seeds germinate in the field and survive to the seedling stage, therefore the quantity of the Z. noltii seeds is not representative of the size of the meadows. The origin of any seeds that germinate also cannot be assumed to be from a specific meadow without genetic analysis, as previous studies in the Wadden Sea display how 70% of the annual recruitment ordinated from outside the existing meadow or were seeds produced within the meadow boundary before 2002.

BI9999 PTY Report C1308785, A1317

55

aerial image dataset provided from the UAV data may also reduce the time-efficiency

of using a drone with the post-processing and generation of an orthomosaic proving

both difficult and time consuming. However, despite these disadvantages the use of

drone-based surveys is still recommended with many future developments into

seagrass UAV monitoring identified from this study (Box 3.0).

Classifications of the aerial data need further development, with the sensitivity

of the mapping often being restricted and not detecting areas low density Z. noltii

(Valle et al. 2015). Biological information such as seagrass morphometric

measurements and above-ground biomass, can also be derived from the NDVI and

Figure 29 Aerial data recorded at Pwllcrochan, July 2016, displaying clear patchiness along the meadow boundary. Strong winds also led to the model (DJI Phantom 3 Professional) blocking part of the photograph. Source: Authors own, 2016.

BI9999 PTY Report C1308785, A1317

56

used to detect changes in Z. noltii distribution. Although the NDVI calculated in this

study is not deemed as true (due to a lack of calibration with the targets present in

the aerial data) values obtained matched those in the literature (Table 15). However,

NDVI must be applied with caution as due to high influences of sediment background

in estuarine conditions (Bargain et al. 2012).

Table 15 The decision rules used to classify Z. noltii beds that had previously been remotely sensed with SPOT multispectral images. Source: Barille et al, 2010.

NDVI Range Description

-1 to 0 Turbid water and bare sediment.

0 - 0.2 Sediment with microphytobenthos and/or low Z. noltii cover.

0.2 – 0.7 Z. noltii with biomass variations

4.3 Conclusion

Z. noltii populations in South and West Wales are currently expanding, but

evidence exists of potentially extensive historic loss throughout the last 50 years,

with changes dependent on both site level environmental factors and anthropogenic

threats. A lack of current consistent monitoring surveys limits knowledge of Z. noltii

distribution. Monitoring programs need to be put in place and should be based on the

methods used in this study. Future UAV seagrass monitoring should take place in

September to encompass the maximum amount of above-ground biomass that can

be sampled and remotely sensed. Seagrass coverage, shoot density and canopy

Box 3.0 Potential developments for the use of a UAV in Z. noltii mapping

The use of a UAV offers the potential for long-term studies to evaluate topographic changes in the seagrass ecosystem. Short term surveys studying seasonality as well as longer time-series analysis could be used to assess how environmental parameters such as elevation change through both orthomosaics and digital elevation models (DEM). Photogrammetric DEMs over time, would allow differences in DEM’s to be computed, illustrating erosion or accretions at specific study sites as displayed by studies into mudflat morphodynamics by Jaud et al (2016). Differences in elevation and topographical features can also be used to enhance orthomosaic classification accuracy (Chust et al. 2010) due to strong association between Z. noltii, estuarine habitat and terrain levels. Furthermore, analysis into the use of a UAV in calculating leaf area index (LAI) would allow calculations into the extent of photosynthesising tissue to take place (Yang and Yang 2009; Wicaksono and Hafizt 2013).

BI9999 PTY Report C1308785, A1317

57

height should be used to monitor the health Z. noltii populations and if financially

viable, parameters such as molar C:N and C:P ratios should be calculated to provide

a true indication of environmental status (Jones and Unsworth 2016). Consistent

monitoring of other intertidal environments such as saltmarsh and mussel crumble

extent will also help to identify the areas that have the potential for Z. noltii

colonisation (Mazik and Boyes 2009; Howson 2012).

5.0 Acknowledgements

First and foremost, I would like to thank all the kind volunteers for all their

assistance with the sampling of this study. I would like to thank Leanne Cullen-

Unsworth, Richard Unsworth, Richard Lilley and Benjamin Jones (Project Seagrass)

for all the guidance and support with my project. I would also like to offer my

gratitude to James P Duffy (University of Exeter) for all the knowledge and

assistance in the field regarding the drone data collection. Finally, I would like to

thank my family for their support.

BI9999 PTY Report C1308785, A1317

58

6.0 References

AgiSoft, L. (2015). Agisoft PhotoScan Professional Edition, Version 1.2.0. http://www.agisoft.com/downloads/installer/.

Alexandre, A., Cabaco, S., Santos, R. and Serrao, E. A. (2006). Timing and success of reproductive stages in the seagrass Zostera noltii. Aquatic Botany 85:219-223.

Anderson, K. and Gaston, K. J. (2013). Lightweight unmanned aerial vechicles will revolutionze spatial ecology. Frontiers in Ecology and the Environment 11:138-146.

Armstrong, R. A. (1993). Remote sensing of submerged vegetation canopies for biomass estimation. Int. J. Remote Sensing 14:621–627.

Bargain, A., Robin, M., Le Men, E., Huete, A. and Barille, L. (2012). Spectral response of the seagrass Zostera noltii with different sediment backgrounds. Aquatic Botany 98:45-56.

Barille, L., Robin, M., Harin, N., Bargain, A. and Launeau, P. (2009). Increase in seagrass distribution at Bourgneuf Bay (France) detected by spatial remote sensing Aquatic Botany.

Bertelli, C. M. and Unsworth, R. K. F. (2014). Protecting the hand that feeds us: Seagrass (Zostera marina) serves as commercial juvenile fish habitat. Marine Polution Bulletin 83:425-429.

Bhattacharya, B., Sarkar, S. K. and Das, R. (2003). Seasonal variations and inherent variability of selenium in marine biota of a tropical wetland ecosystem: implications for bioindicator species. Ecological Indicators 2:367-375.

Borum, J. and Greve, T. M. (2004). The four European seagrass species. In: Borum, J. and Duarte, C.M. and Krause-Jensen, D. and Greve, T.M. (eds.) European Seagrasses: An Introduction to Monitoring and Management. . The M&MS Project, pp. 1-7.

Brown, G. D. (2015). Modelling the potential distribution of Zostera marina in Wales. Degree of Master of Science in Environmental Biology; Conservation and Resource management (MSc) Thesis, Swansea University 2015.

Bunker, F. S. D. (2008). Intertidal SAC monitoring Zostera noltii in Angle Bay, Pembrokeshire Marine SAC 2008 CCW Marine Monitoring Report No: 81, 51pp + ix, Countryside Council for Wales, Bangor.

Carr, M. H., Neigel, J. E., Estes, J. A., Andelman, S., Warner, R. R. and Largier, J. L. (2003). Comparing marine and terrestrial ecosystems: implications for the design of coastal marine reserves. Ecological Applications 13:S90-S107.

Casella, E., Rovere, A., Pedroncini, A., Stark, C. P., Casella, M. and Ferrari, M. (2016). Drones as tools for monitoring beach topography changes in the Ligurian Sea (NW Mediterranean). Geo-Marine Letters 36:151-163.

BI9999 PTY Report C1308785, A1317

59

Charpentier, A., Grillas, P., Lescuyer, F., Coulet, E. and Auby, I. (2005). Spatio-temporal dynamics of a Zostera noltii dominated community over a period of fluctuating salinity in a shallow lagoon, Southern France Estuarine, Coastal and Shelf Science 64:307-315.

Chust, G., Grande, M., Galparsoro, I., Uriarte, A. and Borja, A. (2010). Capabilities of the bathymetric Hawk Eye LiDAR for coastal habitat mapping: a case study within a Basque estuary. Estuarine Coastal and Shelf Science 78:633-643.

Collier, C. J. and Waycott, M. (2014). Temperature extremes reduce seagrass growth and induce mortality. Marine Pollution Bulletin 83:483-490.

Coyer, J. A., Hoarau, G., Kuo, J., Tronholm, A., Veldsink, J. and Olsen, J. L. (2013). Phylogeny and temporal divergence of the seagrass family zosteraceae using one nuclear and three chloroplast loci. Systematics and Biodiversity 11.

Cunha, A. H. and Araujo, A. (2009). New distribution limits of seagrass beds in West Africa. Journal of Biogeography 36:1621-1622.

Davison, D. M. and Hughes, D. J. (1998). Zostera Biotopes (volume I) An overview of dynamics and sensitivity characteristics for conservation management of marine SACs. Scottish Association for Marine Science (UK Marine SACs Project).

den Hartog, C. (1970). The sea-grasses of the World. Verhandelingen Der Koninklijke Nederlandse Akademie Van Wetenschappen, AFD. Natuurkunde 59.

Dolch, T. and Reise, K. (2010). Long-term displacement of intertidal seagrass and mussel beds by expanding sandy bedforms in the northern Wadden Sea. Journal of Sea Research 63:93-101.

Duffy, J. P. and Anderson, K. (2016). A 21st-century renaissance of kites as platforms for proximal sensing. Progress in Physical Geography 40:352-361.

Dugdale, S. J. (2007). An evaluation of imagery from an unmanned aerial vechicle (UAV) for the mapping of intertidal macroalgae on Seal Sands, Tees Estuary, UK, Durham theses, Durham University. Available at Durham E-Theses Online: http://etheses.dur.ac.uk/2879/.

Duggan-Edwards, M. and Brazier, D. P. (2015). Intertidal SAC monitoring Zostera noltii in Angle Bay, Pembrokeshire Marine SAC 2013 NRW Evidence Report No: 55, 38pp +xi, Natural Resources Wales, Bangor

Edwards, R. and White, I. (1998). The Sea Empress oil spill. Terance Dalton, Lavenham, Suffolk.

Grech, A. and Coles, R. G. (2010). An ecosystem-scale predictive model of coastal seagrass distribution. Aquatic Conservation: Marine and Freshwater Ecosystems 20:437-444.

BI9999 PTY Report C1308785, A1317

60

Green, E. P. and Short, F. T. (2003). World Atlas of Seagrasses (Copub: Unep-Wcmc). University of California Press.

Harrison, P. G. (1993). Variations in demography of Zostera marina and Z. noltii on an intertidal gradient. Aquatic Botany 45:63-77.

Hodges, J. and Howe, M. (2007). Milford Haven Waterway Monitoring of Eelgrass, Zostera noltei, following the Sea Empress Oil Spill.

Hootsmans, M. J. M., Vermaat, J. E. and Van Vierssen, W. (1987). Seed-bank development, germination and early seedling survival of two seagrass species from The Netherlands: Zostera marina L. and Zostera noltii hornem. Aquatic Botany 28:275-285.

Howson, C. M. (2012). Intertidal SAC monitoring, Carmarthen Bay SAC, September 2009. CCW Marine Monitoring Report No. 79, 147pp + x, Countryside Council for Wales, Bangor. .

IPCC. (2007). Climate Change 2007: The Physical Science Basis. Summary for Policymakers: 1-21.

Jackson, E. L., Higgs, S., Allsop, T., A, C., Evans, J. and Langmead, O. (2011). Isles of Scilly Seagrass Mapping. Natural England Commissioned Reports, Number 087.

Jaud, M., Grasso, F., Le Dantec, N., Verney, R., Delacourt, C., Ammann, J., Deloffre, J. et al. (2016). Potential of UAV's for Monitoring Mudflat Morphodynamics (Application to the Seine Estuary, France). ISPRS Int. J. Geo-Inf. 5:20.

Johnson, L. and Gage, S. (1997). Landscape approaches to the analysis of aquatic ecosystems. Freshwater Biology 37:113-132.

Jones, B. L. and Unsworth, R. K. F. (2016). The perilous state of seagrass in the British Isles. R. Soc. open sci 3.

Kay, Q. (1998). A review of the existing state of knowledge of the ecology and distribution of seagrass beds around the coast of Wales. Report to the Countryside Council for Wales.

Knudby, A. and Nordlund, L. (2011). Remote sensing of seagrasses in a patchy multi-species environment. International Journal of Remote Sensing 32.

Koedsin, W., Intararuang, W., Ritchie, R. J. and Huete, A. (2016). An Integrated Field and Remote Sensing Method for Mapping Seagrass Species, Cover, and Biomass in Southern Thailand Remote Sensing 8.

Krause-Jensen, D., Quaresma, A. L., Cunha, A. H. and Greve, T. M. (2004). How are seagrass distribution and abundance monitored? In: Borum, J. and Duarte, C.M. and

BI9999 PTY Report C1308785, A1317

61

Krause-Jensen, D. (eds.) European Seagrasses: An introduction to Monitoring and Management. pp. 45-53.

Lomax, A. S., Corso, W. and Etro, J. F. (2005). Employing Unmanned Aerial Vehicles (UAVs) as an Element of the Integrated Ocean Observing System;. Proceedings of the Marine Technology Society/IEEE 1:184-190.

Loques, F., Caye, G. and Meinesz, A. (1988). Flowering and fruiting of Zostera noltii in Golfe Juan (French Mediterranean). Aquatic Botany 32:341-352.

Lyons, M., Phinn, S. and Roelfsema, C. (2011). Integrating Quickbird Multi-Spectral Satellite and Field Data: Mapping Bathymetry, Seagrass Cover, Seagrass Species and Change in Moreton Bay, Australia in 2004 and 2007. Remote Sensing 3:42-64.

Maddock, A. (2011). UK Biodiversity Action Plan Priority Habitat Descriptions. Biodiveristy Reporting and Information Group (BRIG), Joint Nature Conservation Committee.

Marsh, J. A., Dennison, W. C. and Alberte, R. S. (1986). Effects of temperature on photosynthesis and respiration in eelgrass (Zostera marina L.). Journal of Experimental Marine Biology and Ecology 101:257-267.

Massa, S. I., Arnaud-Haond, S., Pearson, G. A. and Serrao, E. A. (2008). Temperature tolerance and survival of intertidal populations of the seagrass Zostera noltii (Hornemann) in Southern Europe (Ria Formosa, Portugal). Hydrobiologia 619:195-201.

Mazik, K. and Boyes, S. (2009). Intertidal monitoring of eelgrass Zostera noltii in the BurryInlet, Carmarthen Bay and Estuaries SAC. CCW Marine Monitoring Report No 53.

McEvoy, J. F., Hall, G. P. and McDonald, P. G. (2016). Evaluation of unmanned aerial vehicle shape, flight path and camera type for waterfowl surveys: disturbance effects and species recognition. PeerJ 4.

McKenzie, L. J., Campbell, S. J. and Roder, C. A. (2003). Methods for Monitoring Seagrass Status.Seagrass-Watch: Manual for Mapping & Monitoring Seagrass Resources by Community (citizen) volunteers. 2nd ed. (QFS, NFC, Cairns), pp. 41-58.

Mckenzie, L. J., Campbell, S. J. and Roder, S. J. (2003). Methods for Monitoring seagrass conditions & resilience.Seagrass-Watch: Manual for Mapping & Monitoring Seagrass Resources by Community (citizen) volunteers. 2nd ed. (QFS, NFC, Cairns), pp. 59-74.

McKenzie, L. J., Collier, C. and Waycott, M. (2012). Reef Rescue Marine Monitoring Program - Inshore Seagrass, Annual Report for the sampling period 1st July 2010 - 31st May 2011. Fisheries Queensland, Cairns. 230pp.

McKenzie, L. J., Collier, C. and Waycott, M. (2014). Reef Resuce Marine Monitoring Program - Inshore Seagrass, Annual Report for the sampling period 1St July 2011 - 31st May 2012. TropWATER, James Cook University, Cairns. 176pp.

BI9999 PTY Report C1308785, A1317

62

Mcleod, E., Chmura, G. L., Bouillon, S., Salm, R., Bjork, M., Duarte, C. M., Lovelock, C. E. et al. (2011). A blueprint for blue carbon: toward an improvement understanding of the role of vegetated coastal habitats in sequestering CO2. Frontiers in Ecology and the Environment 9:552-560.

McMahon, K., van Dijk, K.-J., Ruiz-Montoya, L., Kendrick, G. A., Krauss, S. L., Waycott, M., Verduin, J. et al. (2014). The movement ecology of seagrasses. Proc. R. Soc. B 281.

Meyer, C. A. (2008). Application of remote sensing methods to assess the spatial extent of the seagrass resource in St. Joseph Sound and Clearwater Harbor, Florida, U.S.A Thesis Geography, University of South Florida.

Moore, J. J. (2006). State of the marine environment in SW Wales, 10 years after the Sea Empress oil spill. A report to the Countryside Council for Wales from Coastal Assessment, Liaison & Monitoring, Cosheston, Pembrokeshire. CCW Marine Monitoring Report No: 21. 30pp.

Natural Resources Wales. (2014). Environmental Pressures on the Milford Haven Waterway, Report Reference A&R/SW/14/1.

Natural Resources Wales. (2016). Review of the Trophic Status of the Milford Haven Waterway.

Nikitik, C. (2015). Pembroke Environmental Monitoring Intertidal Eelgrass Report 2014, Report No: JUKL/B1810700/R29.

Orth, R. J., Luckenbach, M. and Moore, K. A. (1994). Seed Dispersal in a Marine Macrophyte: Implications for Colonization and Restoration. Ecology 75:1927-1939.

OSPAR. (2009). Background Document for Zostera beds, Seagrass beds.

Paul, M. (2011). The role of Zostera noltii in wave attenuation. Thesis University of Southampton.

Paul, M. and Amos, C. L. (2011). Spatial and seasonal variation in wave attenuation over Zostera noltii Journal of Geophysical Research 116.

Pergent-Martini, C. and Pergent, G. (2000). Are marine phanerograms a valuable tool in the evaluation of marine trace-metal contamination: example of the Mediterranean sea? International Journal of Environmental Pollution 13:126-147.

Peterson, D. L., Brass, J. A., Smith, W. H., Langford, G., Wegener, S., Dunagan, S., Hammer, P. et al. (2003). Platform options of free-flying satellites, UAVs orthe International Space Station for remote sensing assessment of the littoral zone. International Journal of Remote Sensing 24:2875-2804.

BI9999 PTY Report C1308785, A1317

63

Philippart, C. J. M. and Dijkema, K. S. (1995). Wax and wane of Zostea noltii Hornem. in the Dutch Wadden Sea. Aquatic Botany 49:255-268.

R Core Team. (2013). R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. URL http://www.R-project.org/.

Rasheed, M. A., Thomas, R., Roelofs, A. J., Neil, K. M. and Kerville, S. P. (2003). Port Curtis and Rodds Bay seagrass and benthic macro-invertebrate community baseline survey. DPI Information Series QI03058 (DPI, Cairns).

Reed, B. J. and Hovel, K. A. (2006). Seagrass habitat disturbance: how loss and fragmentation of eelgrass Zostera marina influences epifaunal abundance and diversity. Marine Ecology Progress Series 326:133-143.

Reise, K. and Kohlus, J. (2008). Seagrass recovery in the Northern Wadden Sea? Helgol Mar Res 62:77–84.

Salita, J. T., Ekau, W. and Saint-Paul, U. (2003). Field evidence on the influence of seagrass landscapes on fish abundance in Bolinao, northern Philippines. Marine Ecology Progress Series 247:183-195.

Seitz, S. M., Curless, B., Diebel, J., Scharstein, D. and Szeliski, R. (2006). A comparison and evaluation of multi-view stereo reconstruction algorithms. Proceedings of the 2006 IEEE Computer Society Conference on Computer Vision and Pattern Recognition. IEEE Computer Society, Washington, DC. pp. 519-528.

Short, F. T., Carruthers, T. J. R., Waycott, M., Kendrick, G. A., Fourqurean, J. W., Callabine, A., Kenworthy, W. J. et al. (2010). Zostera noltii. The IUCN Red List of Threatened Species 2010: e.T173361A6999224. http://dx.doi.org/10.2305/IUCN.UK.2010-3.RLTS.T173361A6999224.en. Downloaded on 01 June 2016.

Short, F. T., Polidoro, D., Livingstone, S. R., Carpenter, K. E., Bandeira, S., Bujang, J. S., Calumpong, H. P. et al. (2011). Extinction risk assessment of the world’s seagrass species. Biological Conservation 144:1961–1971.

Stace, C. A. (1997). New Flora of the British Isles. Cambridge: Cambridge University Press.

Telesca, L., Belluscio, A., Criscoli, A., Ardizzone, G., Apostolaki, E. T., Fraschetti, S., Gristina, M. et al. (2015). Seagrass meadows (Posidonia oceanica) distribution and trajectories of change. Sci Rep 5.

Tucker, C. J. (1979). Red and photographic infrared linear combinations for monitoring vegetation. Remote Sensing of Environment 8:127-150.

Uhrin, A. V. and Townsend, P. A. (2016). Improved seagrass mapping using linear spectral unmixing of aerial photographs. Estuarine, Coastal and Shelf Science 171:11-22.

BI9999 PTY Report C1308785, A1317

64

Ullman, S. (1979). The interpretation of structure from motion. Proceedings of the Royal Society B 203:405-426.

Unsworth, R. K. F., Collier, C. J., Waycott, M., Mckenzie, L. J. and Cullen-Unsworth, L. C. (2015). A framework for the resilience of seagrass ecosystems. Marine Pollution Bulletin 13.

Valle, M., Chust, G., Del Campo, A., Wisz, M. S., Olsen, S. M., Gardmendia, J. M. and Borja, A. (2014). Projecting future distribution of the seagrass Zostera noltii under global warming and sea level rise. Biological Conservation 170:74-85.

Valle, M., Pala, V., Lafon, V., Dehouck, A., Garmendia, J. M., Borja, A. and Chust, G. (2015). Mapping estuarine habitats using airborne hyperspectral imagery, with special focus on seagrass meadows. Estuarine, Coastal and Shelf Science 164:433-442.

Ventura, D., Bruno, M., Lasinio, G. J., Belluscio, A. and Ardizzone, G. (2016). A low-cost drone based application for identifying and mapping of coastal fish nursery grounds. Estuarine, Coastal and Shelf Science 171:85-98.

Waycott, M., Duarte, C. M., Carruthers, T. J. B., Orth, R. J. O., Dennison, W. C., Olyarnik, S., Calladine, A. et al. (2009). Accelerating loss of seagrasses across the globe threatens coastal ecosystems. PNAS 106:12377–12381.

Wicaksono, P. and Hafizt, M. (2013). Mapping Seagrass from Space: Addressing the Complexity of Seagrass LAI Mapping. European Journal of Remote Sensing 46:18-39.

Wilde, J. (2016). Influence of Habitat Health on Abundance and Diversity of Juvenile Fish Within Meadows of Posidonia oceanica and its implications for food security in the Eastern Aegean. Dissertation Cardiff University.

WWF. (2012). Historical changes in Welsh Seas: ecosystem trends. Online: In: CYMRU, W (ed.).

Yang, D. and Yang, C. (2009). Detection of Seagrass Distribution Changes from 1991 to 2006 in Xincun Bay, Hainan, with Satellite Remote Sensing. Sensors 9:830-844.

Zar, J. H. (1984). Biostatistical Analysis. Sydney, Australia:Prentice Hall.

Zipperle, A. M., Coyer, J. A., Reise, K., Stam, W. T. and Olsen, J. L. (2009). Evidence for the persistent seed banks in dwraf eelgrass Zostera noltii in the German Wadden Sea. Marine Ecology progress Series 380:73-80.

BI9999 PTY Report C1308785, A1317

65

Appendix 1. Differences in the overlap and computed camera positions of images captured at Angle Bay (July 2016) on an A) modified Xiaomi Yi Action and B) RICOH GR II consumer grade compact camera. Orthomosaics generated using Agisoft Photoscan Professional 1.2.0.

A

9 8 7 6 5 4 3 2 1

9

8

7

6

5

4

3

2

1

B

BI9999 PTY Report C1308785, A1317

66

Appendix 2. Image classification methods of UAV data

Unsupervised classifications

Referenced images were classified using an Iterative Self-Organising Data Analysis Technique (ISODATA). It is based on a k-means clustering algorithm, using multiple iterations to determine data grouping (Jensen, 2005) and no input is required from the surveyor. It was used as an initial investigation to examine the number of significant classes in the composite image (bands 1-3), with the number of classes initially set larger than expected (30). The classes were then reduced through aggregation until up to 10 target classes remained (Z. noltii cover (up to 3 levels), surface water, mud, rock, gravel, trees, fine algae, Ulva spp) depending on the individual study sites.

Supervised classifications

Supervised classifications use the knowledge of the surveyor to define spectral groups. Training areas were selected initially from the composite image using locations sampled during morphometric fieldwork, with extra areas added to ensure a wide variety of pixel types, locations and sizes were included, to generate a training signature. Images were then classified using the parametric (Jensen, 2005) Maximum Likelihood classification (MLC) which assess the variance of the training data with pixel brightness values (Meyer 2008). Points were grouped into similar target classes used in the unsupervised classification; Z. noltii (2 levels of coverage), microphytobenthos (up to 2 levels of coverage), rock/gravel, surface water and mud, that were all identified through ground truthing.

Accuracy Assessments

Post-processing, accuracy assessments were performed on both the unsupervised and supervised raster layers to determine the success of the classification process. 30 random points were tested (to reflect the seagrass morphometric sampling), with the classification from the raw orthomosaic and unsupervised/supervised image analysis being compared. The overall accuracy was calculated from the error matrix generated and Cohen’s Kappa calculation was then used to determine whether or not the results achieved were better than those strictly achieved by chance;

K = PO – PE

1 - PE

O = observed E = expected

BI9999 PTY Report C1308785, A1317

67

Appendix 3. Error Matrices generated from Unsupervised ISO Cluster classifications of true-colour mosaics at a) Pwllcrochan and b) Garren Pill.

Pwllcrochan

Rock/Gravel Surface Water

Dense Z. noltii

Medium density Z. noltii

Sparse Z. noltii/microphytobenthos

Mud

Rock/Gravel 0 0 0 0 0 1

Surface Water 0 1 0 0 0 0

Dense Z. noltii 0 0 1 0 0 0

Medium density Z. noltii 0 0 1 5 2 0

Sparse Z. noltii/microphytobenthos

0 0 2 3 6 1

Mud 0 2 0 0 2 2

Overall Accuracy = 50%

Kappa coefficient = 0.33

Garren Pill

Dense Z. noltii

Trees Medium Z. noltii

Sparse Z. noltii/microphytobenthos

Surface Water

Mud Ulva spp

Dense Z. noltii 5 0 0 0 0 0 0

Trees 0 0 0 0 0 0 0

Medium Z. noltii 1 0 2 0 0 0 0

Sparse Z. noltii/microphytobenthos

2 0 1 0 0 1 0

Surface Water 0 0 0 0 2 1 0

Mud 0 0 2 1 4 8 0

Ulva spp 0 0 0 0 0 0 0

Overall Accuracy = 56.7%

Kappa Coefficient = 0.42

BI9999 PTY Report C1308785, A1317

68

Appendix 4. Error Matrices generated from Supervised Maximum Likelihood image classifications (MLC) of true-colour mosaics at a) Pwllcrochan and b) Garren Pill.

Pwllcrochan

Dense Z. noltii

Medium Dense Z. noltii

Sparse Z. noltii

Mud Dense Microphytobenthos

Sparse Microphytobenthos

Rock/Gravel Surface Water

Dense Z. noltii 3 3 0 0 0 0 0 0

Medium Dense Z. noltii

2 6 0 0 0 0 0 0

Sparse Z. noltii 0 2 0 1 0 0 0 0

Mud 0 0 0 7 0 0 0 0

Dense Microphytobenthos

0 0 0 0 0 1 0 0

Sparse Microphytobenthos

0 0 0 0 0 1 0 0

Rock/Gravel 0 0 0 2 0 0 1 0

Surface Water 0 0 0 0 0 0 0 0

Overall Accuracy = 60%

Kappa Coefficient = 0.48

Garren Pill

Ulva spp

Surface Water

Microphytobenthos Z. noltii Dense Z. noltii

Mud

Ulva spp 0 0 0 0 0 0

Surface Water 0 0 0 0 0 0

Microphytobenthos 0 0 0 0 0 1

Z. noltii 0 0 1 0 2 0

Dense Z. noltii 0 1 0 1 8 0

Mud 0 2 1 3 0 12

Overall Accuracy = 66.7%

Kappa Coefficient = 0.50

BI9999 PTY Report C1308785, A1317

69

Appendix 5. Relationship between the seagrass morphometric values and spatial heterogeneity across the Z. noltii meadows. For Study sites LS1, LS2, LS3 and SP changes in habitat characteristics were not measured, due to the extremely small size of the site (<1ha).

Site: PE

Legend

= Z. noltii % cover = Seed density = Canopy height

= Meadow boundary

Coordinate System: British National GridProjection: Transverse MercatorDatum: OSGB 1936False Easting: 400,000.0000False Northing: -100,000.0000Central Meridian: -2.0000Scale Factor: 0.9996Latitude Of Origin: 49.0000Units: Meter

¯0 0.025 0.05 0.075 0.10.0125

Miles

Legend

Pencla%CoverMap

F__Cover

0.000000 - 12.200000

12.200001 - 24.400000

24.400001 - 36.600000

36.600001 - 48.800000

48.800001 - 61.000000

Coordinate System: British National GridProjection: Transverse MercatorDatum: OSGB 1936False Easting: 400,000.0000False Northing: -100,000.0000Central Meridian: -2.0000Scale Factor: 0.9996Latitude Of Origin: 49.0000Units: Meter

¯0 0.025 0.05 0.075 0.10.0125

Miles

Legend

Pencla%CoverMap

Shoot_Dens

0.000000 - 24.000000

24.000001 - 48.000000

48.000001 - 72.000000

72.000001 - 96.000000

96.000001 - 120.000000

Coordinate System: British National GridProjection: Transverse MercatorDatum: OSGB 1936False Easting: 400,000.0000False Northing: -100,000.0000Central Meridian: -2.0000Scale Factor: 0.9996Latitude Of Origin: 49.0000Units: Meter

¯0 0.025 0.05 0.075 0.10.0125

Miles

Legend

Pencla%CoverMap

Average_Ca

0.000000 - 14.666667

14.666668 - 29.333333

29.333334 - 44.000000

44.000001 - 58.666666

58.666667 - 73.333333

Coordinate System: British National Grid

Projection: Transverse MercatorDatum: OSGB 1936False Easting: 400,000.0000False Northing: -100,000.0000Central Meridian: -2.0000Scale Factor: 0.9996Latitude Of Origin: 49.0000Units: Meter

¯0 0.025 0.05 0.075 0.10.0125

Miles

Legend

Pencla%CoverMap

Average_Ca

0.000000 - 14.666667

14.666668 - 29.333333

29.333334 - 44.000000

44.000001 - 58.666666

58.666667 - 73.333333

BI9999 PTY Report C1308785, A1317

70

Site: AB

Legend

Export_Output

F__Cover

0.000000 - 18.400000

18.400001 - 36.800000

36.800001 - 55.200000

55.200001 - 73.600000

73.600001 - 92.000000

Legend

Export_Output

Average_Le

0.000000 - 14.400000

14.400001 - 28.800000

28.800001 - 43.200000

43.200001 - 57.600000

57.600001 - 72.000000

Legend

Export_Output

Shoot_Dens

0.000000 - 30.200000

30.200001 - 60.400000

60.400001 - 90.600000

90.600001 - 120.800000

120.800001 - 151.000000

Legend

Export_Output_2

Shoot_Dens

0.000000 - 42.000000

42.000001 - 84.000000

84.000001 - 126.000000

126.000001 - 168.000000

168.000001 - 210.000000

Coordinate System: British National GridProjection: Transverse MercatorDatum: OSGB 1936False Easting: 400,000.0000False Northing: -100,000.0000Central Meridian: -2.0000Scale Factor: 0.9996Latitude Of Origin: 49.0000Units: Meter

0 0.015 0.03 0.045 0.060.0075Miles¯

Legend

= Z. noltii % cover = Seed density = Canopy height

= Meadow boundary

BI9999 PTY Report C1308785, A1317

71

Site: PW

Legend

Pwllcrochan Random Point Sampling

Average__

0.000000 - 18.700000

18.700001 - 37.400000

37.400001 - 56.100000

56.100001 - 74.800000

74.800001 - 93.500000

0 0.015 0.03 0.045 0.060.0075Miles

Coordinate System: British National GridProjection: Transverse MercatorDatum: OSGB 1936False Easting: 400,000.0000False Northing: -100,000.0000Central Meridian: -2.0000Scale Factor: 0.9996Latitude Of Origin: 49.0000Units: Meter

¯

Legend

Pwllcrochan Random Point Sampling

Shoot_Dens

0.000000 - 42.000000

42.000001 - 84.000000

84.000001 - 126.000000

126.000001 - 168.000000

168.000001 - 210.000000

0 0.015 0.03 0.045 0.060.0075Miles

Coordinate System: British National GridProjection: Transverse MercatorDatum: OSGB 1936False Easting: 400,000.0000False Northing: -100,000.0000Central Meridian: -2.0000Scale Factor: 0.9996Latitude Of Origin: 49.0000Units: Meter

¯

Legend

Pwllcrochan Random Point Sampling

Average_Ca

0.000000 - 19.266667

19.266668 - 38.533333

38.533334 - 57.800000

57.800001 - 77.066666

77.066667 - 96.333333

0 0.015 0.03 0.045 0.060.0075Miles

Coordinate System: British National GridProjection: Transverse MercatorDatum: OSGB 1936False Easting: 400,000.0000False Northing: -100,000.0000Central Meridian: -2.0000Scale Factor: 0.9996Latitude Of Origin: 49.0000Units: Meter

¯

Legend

Pwllcrochan Random Point Sampling

Average_Ca

0.000000 - 19.266667

19.266668 - 38.533333

38.533334 - 57.800000

57.800001 - 77.066666

77.066667 - 96.333333

0 0.015 0.03 0.045 0.060.0075Miles

Coordinate System: British National GridProjection: Transverse MercatorDatum: OSGB 1936False Easting: 400,000.0000False Northing: -100,000.0000Central Meridian: -2.0000Scale Factor: 0.9996Latitude Of Origin: 49.0000Units: Meter

¯Legend

= Z. noltii % cover = Seed density = Canopy height

= Meadow boundary

BI9999 PTY Report C1308785, A1317

72

Site: PD

Legend

Export_Output_2

Average_Ca

34.000000 - 52.866667

52.866668 - 71.733333

71.733334 - 90.600000

90.600001 - 109.466666

109.466667 - 128.333333

Coordinate System: British National GridProjection: Transverse MercatorDatum: OSGB 1936False Easting: 400,000.0000False Northing: -100,000.0000Central Meridian: -2.0000Scale Factor: 0.9996Latitude Of Origin: 49.0000Units: Meter

0 0.015 0.03 0.045 0.060.0075Miles¯

Legend

Export_Output_2

Average__

3.000000 - 22.000000

22.000001 - 41.000000

41.000001 - 60.000000

60.000001 - 79.000000

79.000001 - 98.000000

Coordinate System: British National GridProjection: Transverse MercatorDatum: OSGB 1936False Easting: 400,000.0000False Northing: -100,000.0000Central Meridian: -2.0000Scale Factor: 0.9996Latitude Of Origin: 49.0000Units: Meter

0 0.015 0.03 0.045 0.060.0075Miles¯

Legend

Export_Output_2

Shoot_Dens

0.000000 - 42.000000

42.000001 - 84.000000

84.000001 - 126.000000

126.000001 - 168.000000

168.000001 - 210.000000

Coordinate System: British National GridProjection: Transverse MercatorDatum: OSGB 1936False Easting: 400,000.0000False Northing: -100,000.0000Central Meridian: -2.0000Scale Factor: 0.9996Latitude Of Origin: 49.0000Units: Meter

0 0.015 0.03 0.045 0.060.0075Miles¯

Legend

Export_Output_2

Shoot_Dens

0.000000 - 42.000000

42.000001 - 84.000000

84.000001 - 126.000000

126.000001 - 168.000000

168.000001 - 210.000000

Coordinate System: British National GridProjection: Transverse MercatorDatum: OSGB 1936False Easting: 400,000.0000False Northing: -100,000.0000Central Meridian: -2.0000Scale Factor: 0.9996Latitude Of Origin: 49.0000Units: Meter

0 0.015 0.03 0.045 0.060.0075Miles¯

Legend

= Z. noltii % cover = Seed density = Canopy height

= Meadow boundary

BI9999 PTY Report C1308785, A1317

73

Site: WW

0 0.05 0.1 0.15 0.20.025Miles¯

Legend

Export_Output

Average_Ca

3.533333 - 4.500000

4.500001 - 6.000000

6.000001 - 7.500000

7.500001 - 9.000000

9.000001 - 10.500000

Coordinate System: British National GridProjection: Transverse MercatorDatum: OSGB 1936False Easting: 400,000.0000False Northing: -100,000.0000Central Meridian: -2.0000Scale Factor: 0.9996Latitude Of Origin: 49.0000Units: Meter

Legend

= Z. noltii % cover = Seed density = Canopy height

= Meadow boundary

Legend

Export_Output

Average__

10.000000 - 25.800000

25.800001 - 41.600000

41.600001 - 57.400000

57.400001 - 73.200000

73.200001 - 89.000000

Legend

Export_Output

Shoot_Dens

25.000000 - 60.000000

60.000001 - 95.000000

95.000001 - 130.000000

130.000001 - 165.000000

165.000001 - 200.000000

Legend

Export_Output

Average_Ca

3.533333 - 4.760000

4.760001 - 5.986667

5.986668 - 7.213333

7.213334 - 8.440000

8.440001 - 9.666667

BI9999 PTY Report C1308785, A1317

74

Site: GP

0 0.05 0.1 0.15 0.20.025Miles¯

Legend

Export_Output

Average__

10.000000 - 25.800000

25.800001 - 41.600000

41.600001 - 57.400000

57.400001 - 73.200000

73.200001 - 89.000000

Coordinate System: British National GridProjection: Transverse MercatorDatum: OSGB 1936False Easting: 400,000.0000False Northing: -100,000.0000Central Meridian: -2.0000Scale Factor: 0.9996Latitude Of Origin: 49.0000Units: Meter

0 0.05 0.1 0.15 0.20.025Miles¯

Legend

Export_Output

Shoot_Dens

25.000000 - 60.000000

60.000001 - 95.000000

95.000001 - 130.000000

130.000001 - 165.000000

165.000001 - 200.000000

Coordinate System: British National GridProjection: Transverse MercatorDatum: OSGB 1936False Easting: 400,000.0000False Northing: -100,000.0000Central Meridian: -2.0000Scale Factor: 0.9996Latitude Of Origin: 49.0000Units: Meter

0 0.05 0.1 0.15 0.20.025Miles¯

Legend

Export_Output

Average_Ca

3.533333 - 4.500000

4.500001 - 6.000000

6.000001 - 7.500000

7.500001 - 9.000000

9.000001 - 10.500000

Coordinate System: British National GridProjection: Transverse MercatorDatum: OSGB 1936False Easting: 400,000.0000False Northing: -100,000.0000Central Meridian: -2.0000Scale Factor: 0.9996Latitude Of Origin: 49.0000Units: Meter

0 0.05 0.1 0.15 0.20.025Miles¯

Legend

Export_Output

Average_Ca

3.533333 - 4.500000

4.500001 - 6.000000

6.000001 - 7.500000

7.500001 - 9.000000

9.000001 - 10.500000

Coordinate System: British National GridProjection: Transverse MercatorDatum: OSGB 1936False Easting: 400,000.0000False Northing: -100,000.0000Central Meridian: -2.0000Scale Factor: 0.9996Latitude Of Origin: 49.0000Units: Meter

Legend

= Z. noltii % cover = Seed density = Canopy height

= Meadow boundary

BI9999 PTY Report C1308785, A1317

75

Appendix 6. Average Principal Components (PC) scores for each of the study sites.

Study Site PC1 PC2

Llanrhidian Sands 1 (LS1) -56.05 5.085667

Llanrhidian Sands 2 (LS2) -52.3467 -10.7057

Llanrhidian Sands 3 (LS3) -60.4 -20.17

Penclacwydd (PE) -21.6323 8.189967

Angle Bay (AB) -2.56933 13.55533

Pwllcrochan (PW) 59.6869 5.862333

Pembroke Dock (PD) 64.91467 -1.335

West Williamson (WW) 75.74333 -4.33413

Garren Pill (GP) 17.89367 -3.56227

Sprinkle Pill (SP) -25.1843 7.466867

Appendix 7. Results of the Z. noltii morphometric GLM

Z. noltii cover (%) Shoot density

(per/0.2m2) Leaf Length (mm)

Intercept 0.0432* 0.909 0.132

Z. noltii Cover X <0.000000225* 0.000132*

Shoot Density <0.000000225* X 0.298

Canopy Height 0.000132* 0.298 X

Seasonality 0.989 0.611 0.527

Sediment

- Mud/Rock 0.0205* 0.733 0.492

- Mud/Sand 0.03498* 0.385 0.588

- Muddy Sand <0.000000225* 0.450 0.608

Elevation 0.915 0.523 0.542

BI9999 PTY Report C1308785, A1317

76

Appendix 8. Relationship between the mean rate of change (μ,yr-1) in seagrass extent and the calculated Zostera Condition Index (ZCI) observed across 10 study sites along the South and West Wales coastline.

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

-30 -20 -10 0 10 20 30 40 50

Zost

era

Co

nd

itio

n In

dex

mean % rate of change, μ (yr-1)

BI9999 PTY Report C1308785, A1317

77

Appendix 9. Random morphometric sampling points superimposed on relative NDVI maps. Maps were generated from fine scale (15m) true-colour and NIR imagery at A) Garren Pill and B) Angle Bay, taken using a modified Xiaomi Yi Action Cameras. Source: L, Pratt, ArcMap 10.2.2, 2016.

A)

B)

BI9999 PTY Report C1308785, A1317

78

Appendix 10. Relationship between Z. noltii seed density and the calculated Zostera Condition Index (ZCI) observed across 10 study sites along the South and West Wales coastline. Appendix 11. Relationship between Z. noltii seed density and the mean rate of change (μ,yr-1) in seagrass extent observed across 10 study sites along the South and West Wales coastline.

-30

-20

-10

0

10

20

30

40

50

0 1 2 3 4 5 6 7

mea

n %

rat

e o

f ch

ange

, μ (

yr-1

)

Seed Density (m2)

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0 1 2 3 4 5 6 7

Zost

era

Co

nd

itio

n In

dex

Seed Density (m2)

BI9999 PTY Report C1308785, A1317

79

Appendix 12. Relationship between Z. noltii seed density and the elevation (above chart datum) of the seagrass meadows.

0

0.5

1

1.5

2

2.5

3

3.5

4

4.5

5

0 1 2 3 4 5 6 7 8 9 10 11 12

Tota

l Nu

mb

er o

f Z.

no

ltii

seed

s fo

un

d

Elevation (m)